Feat(ingestion): add canvas pipeline debug (dry-run) mode with View result log (#17538)
## Summary
Adds a side-effect-free DataFlow canvas **debug (dry-run) mode** plus a
**debug run log with a "View result" panel**, so a canvas can be
executed synchronously and inspected end-to-end (per-component progress
and parsed chunks) without persisting anything.
### Dry-run execution (inline parsed chunks)
- `task/debug.go`: `NewDebugTaskContext` builds an in-memory
`TaskContext` with **`KB.ID == ""` — the single debug signal used across
the ingestion pipeline**. A canvas debug run has no knowledgebase, and
production ingestion always supplies one, so `kb_id == ""` occurs ONLY
in debug mode. Components gate their own side effects on this signal
without any dedicated debug vocabulary (the former `CANVAS_DEBUG_DOC_ID`
marker constant is removed).
- `task/pipeline_executor.go`: `validateTaskContext` no longer requires
a KB when `KB.ID == ""` (debug); debug runs return `collectDebugOutput`
(chunks) instead of a no-op; uploaded bytes are delivered as
`inputs['binary']` for doc-less runs; `injectDebugPageCap` caps the
parser to the first pages for a fast preview via the production
`override_params` channel (Parser cpnID + family).
- `component/tokenizer.go`: `shouldHaveEmbedding` skips embedding when
`kb_id == ""` — the embedder is configured on the knowledgebase, so a
debug run has nothing to resolve against and stays side-effect free.
- `chunker/register.go`: chunk images are uploaded to MinIO only when a
KB is present (persist run). **In debug mode the raw image bytes are
intentionally dropped (`delete(ck, "image")`)** — the debug preview does
not render chunk images, and dropping the bytes keeps them out of memory
and out of the Redis-stored debug log. This is a deliberate trade-off,
not an oversight.
- `component/file.go`: pass through in-memory binary bytes, skipping
`doc_id` -> storage resolution.
- `handler/agent.go` + `agent_webhook.go`: detect `dataflow_canvas` and
run a sync debug returning chunks inline on the existing
chat/completions endpoint; reject DataFlow canvases from webhooks (fixes
the previously dead `== "DataFlow"` check; mirrors Python
`agent_api.py`).
- `parser_dispatch.go`: export `ParserFileFamily` for the executor's
page-cap injection.
### Debug run log + "View result"
Mirrors Python's debug-log contract so the front-end can replay each
component's progress and parsed output:
- `task/debug_log_sink.go`: a `DebugLogSink` records every component's
lifecycle into a `[{component_id, trace}]` array (each trace entry
carries `message`, `progress`, `timestamp`, `elapsed_time`). `Flush`
appends a terminal `END` marker whose first trace message is non-empty
so the front-end detects completion. On failure the END marker is
prefixed `[ERROR]` yet still carries the run, so the failure timeline
renders instead of being stuck empty. Timestamps and `elapsed_time` are
in seconds (matching the rest of the app).
- `task/debug_result_dsl.go`: `BuildDebugResultDSL` builds the `dsl` the
END marker carries — the Go analogue of Python's `Graph.__str__` +
END-marker `dsl` in `rag/flow/pipeline.py`. It combines the static DSL
structure (component_name / downstream / params / graph.nodes) with the
run output map (`output["state"][<id>]`) to emit, per component,
`obj.params.outputs[<format>].value` (chunks / text / json / html /
markdown) — the exact keys the front-end `dataflow-result` page reads to
render each step's parsed chunks. Raw embedding vectors (including the
dimension-scoped `q_<dim>_vec` keys) are stripped so the stored log
stays Python-scale.
- `task/pipeline_executor.go`: after the run, attach the built `dsl` to
the END marker via the `ResultSink` capability.
- `handler/agent.go`: `runCanvasPipelineDebug` generates a stable
`message_id` up-front and always flushes the log (success or failure);
`respondWithDebugResult` returns `message_id` in **both** the success
and the error envelope so the front-end can poll the log. The debug-log
endpoint `GET /agents/:id/logs/:message_id` serves the array.
- `web/src/pages/agent/hooks/use-run-dataflow.ts`: on a run failure,
also surface `message_id` via `setMessageId` so the log sheet renders
the failure timeline (the `[ERROR]` END marker is already written).
Guarded by `if (msgId)`, so it is a safe no-op when the back-end does
not return an id.
## Behavioral notes
- Debug parses only the first pages (`debugPageCapPages`) for a fast
preview; an explicit `pages` cap already present in the ParserConfig is
respected.
- Debug mode does not keep chunk images (see above) and does not compute
embeddings — it exercises parse + chunk only.
## Test plan
- Go: `debug_test.go`, `debug_log_sink_test.go` (trace pairing, END
marker, `[ERROR]` prefix, fractional-second timestamp/elapsed_time, size
caps, and a real-pipeline test asserting the END-marker `dsl` carries
non-empty per-component `params.outputs` with chunks),
`debug_result_dsl_test.go` (flat and real nested `output["state"]`
shapes, vector stripping, format priority),
`debug_pages_integration_test.go`, `pipeline_executor_persist_test.go`,
`handler/agent_pipeline_debug_test.go`, `handler/agent_logs_test.go`
(incl. `TestRunCanvasPipelineDebug_ErrorStillExposesMessageID` /
`TestRespondWithDebugResult_ErrorCarriesMessageID` locking `message_id`
on failure), plus updates to `agent_test.go` / `agent_webhook_test.go` /
`chunker/image_upload_test.go` / `tokenizer*.go`.
- `go build ./...` and `./build.sh --test` for affected packages.
🤖 Generated with [CodeBuddy Code](https://cnb.cool/codebuddy)
---------
Co-authored-by: CodeBuddy Code <noreply@tencent.com>
2026-07-31 13:10:08 +08:00
//
// Copyright 2026 The InfiniFlow Authors. All Rights Reserved.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
//
package task
import (
"context"
"encoding/json"
"fmt"
"os"
"path/filepath"
"reflect"
refactor(task): sink parser page-cap override into pipeline package (#17905)
## Summary
Moves the canvas-debug parser page-cap injection out of the `task`
orchestrator and into a **debug-agnostic** `pipeline` helper, so
`PipelineExecutor` keeps only the orchestration skeleton (resolving one
of the P1 review findings: the executor was overloaded with
DSL/parser-param assembly).
### Changes
- **`pipeline/parser_page_cap.go`** (new):
- `BuildParserPageCapOverride(parserConfig, dsl, docType, capPages int,
parserComponentName string, familyOf)` — injects the
`ParserConfig[cpnID][family]["pages"]` cap through the same
`override_params` channel production uses. The cap value and family
resolution are injected by the caller, so the function carries no debug
semantics and is reusable for any page-cap scenario.
- `ExtractParserCpnID(dsl, parserComponentName)` — shared Parser cpnID
discovery from (optionally enveloped) DSL.
- `UnwrapCanvasDSL(raw []byte)` — exported single source of truth for
stripping the `{"dsl": {...}}` canvas envelope.
- `pipeline` does **not** import `component` (no reverse dependency);
callers inject `component.ComponentNameParser` /
`component.ParserFileFamily`.
- **`task/pipeline_executor.go`**: removed `injectDebugPageCap` (the
`debugPageCapPages = 2` constant stays in the task package). The debug
branch now calls `pipeline.BuildParserPageCapOverride(...)`.
- **`task/pipeline_executor.go` `warnUnknownComponentParams`**: fixed a
production no-op bug — it passed the enveloped DSL straight to
`ExtractAllComponentParams`, which silently errored and disabled the
unknown-cpnID guard. It now unwraps the envelope first.
- **`task/debug_result_dsl.go`**: reuses `pipeline.UnwrapCanvasDSL`
instead of a third inline envelope-unwrap copy.
### Behavior
No external debug-preview behavior changes. The three original
invariants are preserved exactly:
1. explicit `pages` caps under `cpnID+family` are respected (not
overwritten),
2. an empty family (unknown docType) is a no-op,
3. the injected shape is `[]any{[]any{1, capPages}}` (the
`[]any`-of-`[]any` form `NormalizePDFPages` requires).
## Test plan
- New `pipeline/parser_page_cap_test.go`: `BuildParserPageCapOverride`
(inject / respect-existing / unknown-family no-op / no-Parser no-op),
`ExtractParserCpnID` (enveloped + raw), `UnwrapCanvasDSL`.
- `task/debug_test.go`: `TestInjectDebugPageCap` migrated to the new
helper; new
`TestWarnUnknownComponentParamsDetectsUnknownCPNFromEnvelope` captures
the warning via `zaptest/observer` to prove the envelope no-op bug is
fixed.
- Both `internal/ingestion/pipeline` and `internal/ingestion/task` pass
`build.sh --test` (unit tier).
## Notes
- `TOKEN_CHUNKER_HANDOFF.md` is an unrelated untracked file and was
deliberately **not** included in this PR.
2026-08-06 15:50:39 +08:00
"strings"
Feat(ingestion): add canvas pipeline debug (dry-run) mode with View result log (#17538)
## Summary
Adds a side-effect-free DataFlow canvas **debug (dry-run) mode** plus a
**debug run log with a "View result" panel**, so a canvas can be
executed synchronously and inspected end-to-end (per-component progress
and parsed chunks) without persisting anything.
### Dry-run execution (inline parsed chunks)
- `task/debug.go`: `NewDebugTaskContext` builds an in-memory
`TaskContext` with **`KB.ID == ""` — the single debug signal used across
the ingestion pipeline**. A canvas debug run has no knowledgebase, and
production ingestion always supplies one, so `kb_id == ""` occurs ONLY
in debug mode. Components gate their own side effects on this signal
without any dedicated debug vocabulary (the former `CANVAS_DEBUG_DOC_ID`
marker constant is removed).
- `task/pipeline_executor.go`: `validateTaskContext` no longer requires
a KB when `KB.ID == ""` (debug); debug runs return `collectDebugOutput`
(chunks) instead of a no-op; uploaded bytes are delivered as
`inputs['binary']` for doc-less runs; `injectDebugPageCap` caps the
parser to the first pages for a fast preview via the production
`override_params` channel (Parser cpnID + family).
- `component/tokenizer.go`: `shouldHaveEmbedding` skips embedding when
`kb_id == ""` — the embedder is configured on the knowledgebase, so a
debug run has nothing to resolve against and stays side-effect free.
- `chunker/register.go`: chunk images are uploaded to MinIO only when a
KB is present (persist run). **In debug mode the raw image bytes are
intentionally dropped (`delete(ck, "image")`)** — the debug preview does
not render chunk images, and dropping the bytes keeps them out of memory
and out of the Redis-stored debug log. This is a deliberate trade-off,
not an oversight.
- `component/file.go`: pass through in-memory binary bytes, skipping
`doc_id` -> storage resolution.
- `handler/agent.go` + `agent_webhook.go`: detect `dataflow_canvas` and
run a sync debug returning chunks inline on the existing
chat/completions endpoint; reject DataFlow canvases from webhooks (fixes
the previously dead `== "DataFlow"` check; mirrors Python
`agent_api.py`).
- `parser_dispatch.go`: export `ParserFileFamily` for the executor's
page-cap injection.
### Debug run log + "View result"
Mirrors Python's debug-log contract so the front-end can replay each
component's progress and parsed output:
- `task/debug_log_sink.go`: a `DebugLogSink` records every component's
lifecycle into a `[{component_id, trace}]` array (each trace entry
carries `message`, `progress`, `timestamp`, `elapsed_time`). `Flush`
appends a terminal `END` marker whose first trace message is non-empty
so the front-end detects completion. On failure the END marker is
prefixed `[ERROR]` yet still carries the run, so the failure timeline
renders instead of being stuck empty. Timestamps and `elapsed_time` are
in seconds (matching the rest of the app).
- `task/debug_result_dsl.go`: `BuildDebugResultDSL` builds the `dsl` the
END marker carries — the Go analogue of Python's `Graph.__str__` +
END-marker `dsl` in `rag/flow/pipeline.py`. It combines the static DSL
structure (component_name / downstream / params / graph.nodes) with the
run output map (`output["state"][<id>]`) to emit, per component,
`obj.params.outputs[<format>].value` (chunks / text / json / html /
markdown) — the exact keys the front-end `dataflow-result` page reads to
render each step's parsed chunks. Raw embedding vectors (including the
dimension-scoped `q_<dim>_vec` keys) are stripped so the stored log
stays Python-scale.
- `task/pipeline_executor.go`: after the run, attach the built `dsl` to
the END marker via the `ResultSink` capability.
- `handler/agent.go`: `runCanvasPipelineDebug` generates a stable
`message_id` up-front and always flushes the log (success or failure);
`respondWithDebugResult` returns `message_id` in **both** the success
and the error envelope so the front-end can poll the log. The debug-log
endpoint `GET /agents/:id/logs/:message_id` serves the array.
- `web/src/pages/agent/hooks/use-run-dataflow.ts`: on a run failure,
also surface `message_id` via `setMessageId` so the log sheet renders
the failure timeline (the `[ERROR]` END marker is already written).
Guarded by `if (msgId)`, so it is a safe no-op when the back-end does
not return an id.
## Behavioral notes
- Debug parses only the first pages (`debugPageCapPages`) for a fast
preview; an explicit `pages` cap already present in the ParserConfig is
respected.
- Debug mode does not keep chunk images (see above) and does not compute
embeddings — it exercises parse + chunk only.
## Test plan
- Go: `debug_test.go`, `debug_log_sink_test.go` (trace pairing, END
marker, `[ERROR]` prefix, fractional-second timestamp/elapsed_time, size
caps, and a real-pipeline test asserting the END-marker `dsl` carries
non-empty per-component `params.outputs` with chunks),
`debug_result_dsl_test.go` (flat and real nested `output["state"]`
shapes, vector stripping, format priority),
`debug_pages_integration_test.go`, `pipeline_executor_persist_test.go`,
`handler/agent_pipeline_debug_test.go`, `handler/agent_logs_test.go`
(incl. `TestRunCanvasPipelineDebug_ErrorStillExposesMessageID` /
`TestRespondWithDebugResult_ErrorCarriesMessageID` locking `message_id`
on failure), plus updates to `agent_test.go` / `agent_webhook_test.go` /
`chunker/image_upload_test.go` / `tokenizer*.go`.
- `go build ./...` and `./build.sh --test` for affected packages.
🤖 Generated with [CodeBuddy Code](https://cnb.cool/codebuddy)
---------
Co-authored-by: CodeBuddy Code <noreply@tencent.com>
2026-07-31 13:10:08 +08:00
"testing"
refactor(task): sink parser page-cap override into pipeline package (#17905)
## Summary
Moves the canvas-debug parser page-cap injection out of the `task`
orchestrator and into a **debug-agnostic** `pipeline` helper, so
`PipelineExecutor` keeps only the orchestration skeleton (resolving one
of the P1 review findings: the executor was overloaded with
DSL/parser-param assembly).
### Changes
- **`pipeline/parser_page_cap.go`** (new):
- `BuildParserPageCapOverride(parserConfig, dsl, docType, capPages int,
parserComponentName string, familyOf)` — injects the
`ParserConfig[cpnID][family]["pages"]` cap through the same
`override_params` channel production uses. The cap value and family
resolution are injected by the caller, so the function carries no debug
semantics and is reusable for any page-cap scenario.
- `ExtractParserCpnID(dsl, parserComponentName)` — shared Parser cpnID
discovery from (optionally enveloped) DSL.
- `UnwrapCanvasDSL(raw []byte)` — exported single source of truth for
stripping the `{"dsl": {...}}` canvas envelope.
- `pipeline` does **not** import `component` (no reverse dependency);
callers inject `component.ComponentNameParser` /
`component.ParserFileFamily`.
- **`task/pipeline_executor.go`**: removed `injectDebugPageCap` (the
`debugPageCapPages = 2` constant stays in the task package). The debug
branch now calls `pipeline.BuildParserPageCapOverride(...)`.
- **`task/pipeline_executor.go` `warnUnknownComponentParams`**: fixed a
production no-op bug — it passed the enveloped DSL straight to
`ExtractAllComponentParams`, which silently errored and disabled the
unknown-cpnID guard. It now unwraps the envelope first.
- **`task/debug_result_dsl.go`**: reuses `pipeline.UnwrapCanvasDSL`
instead of a third inline envelope-unwrap copy.
### Behavior
No external debug-preview behavior changes. The three original
invariants are preserved exactly:
1. explicit `pages` caps under `cpnID+family` are respected (not
overwritten),
2. an empty family (unknown docType) is a no-op,
3. the injected shape is `[]any{[]any{1, capPages}}` (the
`[]any`-of-`[]any` form `NormalizePDFPages` requires).
## Test plan
- New `pipeline/parser_page_cap_test.go`: `BuildParserPageCapOverride`
(inject / respect-existing / unknown-family no-op / no-Parser no-op),
`ExtractParserCpnID` (enveloped + raw), `UnwrapCanvasDSL`.
- `task/debug_test.go`: `TestInjectDebugPageCap` migrated to the new
helper; new
`TestWarnUnknownComponentParamsDetectsUnknownCPNFromEnvelope` captures
the warning via `zaptest/observer` to prove the envelope no-op bug is
fixed.
- Both `internal/ingestion/pipeline` and `internal/ingestion/task` pass
`build.sh --test` (unit tier).
## Notes
- `TOKEN_CHUNKER_HANDOFF.md` is an unrelated untracked file and was
deliberately **not** included in this PR.
2026-08-06 15:50:39 +08:00
"go.uber.org/zap"
"go.uber.org/zap/zaptest/observer"
Feat(ingestion): add canvas pipeline debug (dry-run) mode with View result log (#17538)
## Summary
Adds a side-effect-free DataFlow canvas **debug (dry-run) mode** plus a
**debug run log with a "View result" panel**, so a canvas can be
executed synchronously and inspected end-to-end (per-component progress
and parsed chunks) without persisting anything.
### Dry-run execution (inline parsed chunks)
- `task/debug.go`: `NewDebugTaskContext` builds an in-memory
`TaskContext` with **`KB.ID == ""` — the single debug signal used across
the ingestion pipeline**. A canvas debug run has no knowledgebase, and
production ingestion always supplies one, so `kb_id == ""` occurs ONLY
in debug mode. Components gate their own side effects on this signal
without any dedicated debug vocabulary (the former `CANVAS_DEBUG_DOC_ID`
marker constant is removed).
- `task/pipeline_executor.go`: `validateTaskContext` no longer requires
a KB when `KB.ID == ""` (debug); debug runs return `collectDebugOutput`
(chunks) instead of a no-op; uploaded bytes are delivered as
`inputs['binary']` for doc-less runs; `injectDebugPageCap` caps the
parser to the first pages for a fast preview via the production
`override_params` channel (Parser cpnID + family).
- `component/tokenizer.go`: `shouldHaveEmbedding` skips embedding when
`kb_id == ""` — the embedder is configured on the knowledgebase, so a
debug run has nothing to resolve against and stays side-effect free.
- `chunker/register.go`: chunk images are uploaded to MinIO only when a
KB is present (persist run). **In debug mode the raw image bytes are
intentionally dropped (`delete(ck, "image")`)** — the debug preview does
not render chunk images, and dropping the bytes keeps them out of memory
and out of the Redis-stored debug log. This is a deliberate trade-off,
not an oversight.
- `component/file.go`: pass through in-memory binary bytes, skipping
`doc_id` -> storage resolution.
- `handler/agent.go` + `agent_webhook.go`: detect `dataflow_canvas` and
run a sync debug returning chunks inline on the existing
chat/completions endpoint; reject DataFlow canvases from webhooks (fixes
the previously dead `== "DataFlow"` check; mirrors Python
`agent_api.py`).
- `parser_dispatch.go`: export `ParserFileFamily` for the executor's
page-cap injection.
### Debug run log + "View result"
Mirrors Python's debug-log contract so the front-end can replay each
component's progress and parsed output:
- `task/debug_log_sink.go`: a `DebugLogSink` records every component's
lifecycle into a `[{component_id, trace}]` array (each trace entry
carries `message`, `progress`, `timestamp`, `elapsed_time`). `Flush`
appends a terminal `END` marker whose first trace message is non-empty
so the front-end detects completion. On failure the END marker is
prefixed `[ERROR]` yet still carries the run, so the failure timeline
renders instead of being stuck empty. Timestamps and `elapsed_time` are
in seconds (matching the rest of the app).
- `task/debug_result_dsl.go`: `BuildDebugResultDSL` builds the `dsl` the
END marker carries — the Go analogue of Python's `Graph.__str__` +
END-marker `dsl` in `rag/flow/pipeline.py`. It combines the static DSL
structure (component_name / downstream / params / graph.nodes) with the
run output map (`output["state"][<id>]`) to emit, per component,
`obj.params.outputs[<format>].value` (chunks / text / json / html /
markdown) — the exact keys the front-end `dataflow-result` page reads to
render each step's parsed chunks. Raw embedding vectors (including the
dimension-scoped `q_<dim>_vec` keys) are stripped so the stored log
stays Python-scale.
- `task/pipeline_executor.go`: after the run, attach the built `dsl` to
the END marker via the `ResultSink` capability.
- `handler/agent.go`: `runCanvasPipelineDebug` generates a stable
`message_id` up-front and always flushes the log (success or failure);
`respondWithDebugResult` returns `message_id` in **both** the success
and the error envelope so the front-end can poll the log. The debug-log
endpoint `GET /agents/:id/logs/:message_id` serves the array.
- `web/src/pages/agent/hooks/use-run-dataflow.ts`: on a run failure,
also surface `message_id` via `setMessageId` so the log sheet renders
the failure timeline (the `[ERROR]` END marker is already written).
Guarded by `if (msgId)`, so it is a safe no-op when the back-end does
not return an id.
## Behavioral notes
- Debug parses only the first pages (`debugPageCapPages`) for a fast
preview; an explicit `pages` cap already present in the ParserConfig is
respected.
- Debug mode does not keep chunk images (see above) and does not compute
embeddings — it exercises parse + chunk only.
## Test plan
- Go: `debug_test.go`, `debug_log_sink_test.go` (trace pairing, END
marker, `[ERROR]` prefix, fractional-second timestamp/elapsed_time, size
caps, and a real-pipeline test asserting the END-marker `dsl` carries
non-empty per-component `params.outputs` with chunks),
`debug_result_dsl_test.go` (flat and real nested `output["state"]`
shapes, vector stripping, format priority),
`debug_pages_integration_test.go`, `pipeline_executor_persist_test.go`,
`handler/agent_pipeline_debug_test.go`, `handler/agent_logs_test.go`
(incl. `TestRunCanvasPipelineDebug_ErrorStillExposesMessageID` /
`TestRespondWithDebugResult_ErrorCarriesMessageID` locking `message_id`
on failure), plus updates to `agent_test.go` / `agent_webhook_test.go` /
`chunker/image_upload_test.go` / `tokenizer*.go`.
- `go build ./...` and `./build.sh --test` for affected packages.
🤖 Generated with [CodeBuddy Code](https://cnb.cool/codebuddy)
---------
Co-authored-by: CodeBuddy Code <noreply@tencent.com>
2026-07-31 13:10:08 +08:00
"gorm.io/gorm"
refactor(task): sink parser page-cap override into pipeline package (#17905)
## Summary
Moves the canvas-debug parser page-cap injection out of the `task`
orchestrator and into a **debug-agnostic** `pipeline` helper, so
`PipelineExecutor` keeps only the orchestration skeleton (resolving one
of the P1 review findings: the executor was overloaded with
DSL/parser-param assembly).
### Changes
- **`pipeline/parser_page_cap.go`** (new):
- `BuildParserPageCapOverride(parserConfig, dsl, docType, capPages int,
parserComponentName string, familyOf)` — injects the
`ParserConfig[cpnID][family]["pages"]` cap through the same
`override_params` channel production uses. The cap value and family
resolution are injected by the caller, so the function carries no debug
semantics and is reusable for any page-cap scenario.
- `ExtractParserCpnID(dsl, parserComponentName)` — shared Parser cpnID
discovery from (optionally enveloped) DSL.
- `UnwrapCanvasDSL(raw []byte)` — exported single source of truth for
stripping the `{"dsl": {...}}` canvas envelope.
- `pipeline` does **not** import `component` (no reverse dependency);
callers inject `component.ComponentNameParser` /
`component.ParserFileFamily`.
- **`task/pipeline_executor.go`**: removed `injectDebugPageCap` (the
`debugPageCapPages = 2` constant stays in the task package). The debug
branch now calls `pipeline.BuildParserPageCapOverride(...)`.
- **`task/pipeline_executor.go` `warnUnknownComponentParams`**: fixed a
production no-op bug — it passed the enveloped DSL straight to
`ExtractAllComponentParams`, which silently errored and disabled the
unknown-cpnID guard. It now unwraps the envelope first.
- **`task/debug_result_dsl.go`**: reuses `pipeline.UnwrapCanvasDSL`
instead of a third inline envelope-unwrap copy.
### Behavior
No external debug-preview behavior changes. The three original
invariants are preserved exactly:
1. explicit `pages` caps under `cpnID+family` are respected (not
overwritten),
2. an empty family (unknown docType) is a no-op,
3. the injected shape is `[]any{[]any{1, capPages}}` (the
`[]any`-of-`[]any` form `NormalizePDFPages` requires).
## Test plan
- New `pipeline/parser_page_cap_test.go`: `BuildParserPageCapOverride`
(inject / respect-existing / unknown-family no-op / no-Parser no-op),
`ExtractParserCpnID` (enveloped + raw), `UnwrapCanvasDSL`.
- `task/debug_test.go`: `TestInjectDebugPageCap` migrated to the new
helper; new
`TestWarnUnknownComponentParamsDetectsUnknownCPNFromEnvelope` captures
the warning via `zaptest/observer` to prove the envelope no-op bug is
fixed.
- Both `internal/ingestion/pipeline` and `internal/ingestion/task` pass
`build.sh --test` (unit tier).
## Notes
- `TOKEN_CHUNKER_HANDOFF.md` is an unrelated untracked file and was
deliberately **not** included in this PR.
2026-08-06 15:50:39 +08:00
"ragflow/internal/common"
Feat(ingestion): add canvas pipeline debug (dry-run) mode with View result log (#17538)
## Summary
Adds a side-effect-free DataFlow canvas **debug (dry-run) mode** plus a
**debug run log with a "View result" panel**, so a canvas can be
executed synchronously and inspected end-to-end (per-component progress
and parsed chunks) without persisting anything.
### Dry-run execution (inline parsed chunks)
- `task/debug.go`: `NewDebugTaskContext` builds an in-memory
`TaskContext` with **`KB.ID == ""` — the single debug signal used across
the ingestion pipeline**. A canvas debug run has no knowledgebase, and
production ingestion always supplies one, so `kb_id == ""` occurs ONLY
in debug mode. Components gate their own side effects on this signal
without any dedicated debug vocabulary (the former `CANVAS_DEBUG_DOC_ID`
marker constant is removed).
- `task/pipeline_executor.go`: `validateTaskContext` no longer requires
a KB when `KB.ID == ""` (debug); debug runs return `collectDebugOutput`
(chunks) instead of a no-op; uploaded bytes are delivered as
`inputs['binary']` for doc-less runs; `injectDebugPageCap` caps the
parser to the first pages for a fast preview via the production
`override_params` channel (Parser cpnID + family).
- `component/tokenizer.go`: `shouldHaveEmbedding` skips embedding when
`kb_id == ""` — the embedder is configured on the knowledgebase, so a
debug run has nothing to resolve against and stays side-effect free.
- `chunker/register.go`: chunk images are uploaded to MinIO only when a
KB is present (persist run). **In debug mode the raw image bytes are
intentionally dropped (`delete(ck, "image")`)** — the debug preview does
not render chunk images, and dropping the bytes keeps them out of memory
and out of the Redis-stored debug log. This is a deliberate trade-off,
not an oversight.
- `component/file.go`: pass through in-memory binary bytes, skipping
`doc_id` -> storage resolution.
- `handler/agent.go` + `agent_webhook.go`: detect `dataflow_canvas` and
run a sync debug returning chunks inline on the existing
chat/completions endpoint; reject DataFlow canvases from webhooks (fixes
the previously dead `== "DataFlow"` check; mirrors Python
`agent_api.py`).
- `parser_dispatch.go`: export `ParserFileFamily` for the executor's
page-cap injection.
### Debug run log + "View result"
Mirrors Python's debug-log contract so the front-end can replay each
component's progress and parsed output:
- `task/debug_log_sink.go`: a `DebugLogSink` records every component's
lifecycle into a `[{component_id, trace}]` array (each trace entry
carries `message`, `progress`, `timestamp`, `elapsed_time`). `Flush`
appends a terminal `END` marker whose first trace message is non-empty
so the front-end detects completion. On failure the END marker is
prefixed `[ERROR]` yet still carries the run, so the failure timeline
renders instead of being stuck empty. Timestamps and `elapsed_time` are
in seconds (matching the rest of the app).
- `task/debug_result_dsl.go`: `BuildDebugResultDSL` builds the `dsl` the
END marker carries — the Go analogue of Python's `Graph.__str__` +
END-marker `dsl` in `rag/flow/pipeline.py`. It combines the static DSL
structure (component_name / downstream / params / graph.nodes) with the
run output map (`output["state"][<id>]`) to emit, per component,
`obj.params.outputs[<format>].value` (chunks / text / json / html /
markdown) — the exact keys the front-end `dataflow-result` page reads to
render each step's parsed chunks. Raw embedding vectors (including the
dimension-scoped `q_<dim>_vec` keys) are stripped so the stored log
stays Python-scale.
- `task/pipeline_executor.go`: after the run, attach the built `dsl` to
the END marker via the `ResultSink` capability.
- `handler/agent.go`: `runCanvasPipelineDebug` generates a stable
`message_id` up-front and always flushes the log (success or failure);
`respondWithDebugResult` returns `message_id` in **both** the success
and the error envelope so the front-end can poll the log. The debug-log
endpoint `GET /agents/:id/logs/:message_id` serves the array.
- `web/src/pages/agent/hooks/use-run-dataflow.ts`: on a run failure,
also surface `message_id` via `setMessageId` so the log sheet renders
the failure timeline (the `[ERROR]` END marker is already written).
Guarded by `if (msgId)`, so it is a safe no-op when the back-end does
not return an id.
## Behavioral notes
- Debug parses only the first pages (`debugPageCapPages`) for a fast
preview; an explicit `pages` cap already present in the ParserConfig is
respected.
- Debug mode does not keep chunk images (see above) and does not compute
embeddings — it exercises parse + chunk only.
## Test plan
- Go: `debug_test.go`, `debug_log_sink_test.go` (trace pairing, END
marker, `[ERROR]` prefix, fractional-second timestamp/elapsed_time, size
caps, and a real-pipeline test asserting the END-marker `dsl` carries
non-empty per-component `params.outputs` with chunks),
`debug_result_dsl_test.go` (flat and real nested `output["state"]`
shapes, vector stripping, format priority),
`debug_pages_integration_test.go`, `pipeline_executor_persist_test.go`,
`handler/agent_pipeline_debug_test.go`, `handler/agent_logs_test.go`
(incl. `TestRunCanvasPipelineDebug_ErrorStillExposesMessageID` /
`TestRespondWithDebugResult_ErrorCarriesMessageID` locking `message_id`
on failure), plus updates to `agent_test.go` / `agent_webhook_test.go` /
`chunker/image_upload_test.go` / `tokenizer*.go`.
- `go build ./...` and `./build.sh --test` for affected packages.
🤖 Generated with [CodeBuddy Code](https://cnb.cool/codebuddy)
---------
Co-authored-by: CodeBuddy Code <noreply@tencent.com>
2026-07-31 13:10:08 +08:00
"ragflow/internal/entity"
"ragflow/internal/ingestion/component"
"ragflow/internal/ingestion/pipeline"
)
// TestNewDebugTaskContext_InjectsDebugID asserts the debug context carries the
// side-effect-free markers: a fresh non-empty Doc.ID (a throwaway uuid, not a
// persisted row) and an empty KB (debug has no knowledgebase, so kb_id == "" is
// the debug signal used across the pipeline). The parser page cap is no longer
// stored as a flat ParserConfig key here — flat keys are dropped by the
// override_params merge and never reach the parser. It is injected at run time
refactor(task): sink parser page-cap override into pipeline package (#17905)
## Summary
Moves the canvas-debug parser page-cap injection out of the `task`
orchestrator and into a **debug-agnostic** `pipeline` helper, so
`PipelineExecutor` keeps only the orchestration skeleton (resolving one
of the P1 review findings: the executor was overloaded with
DSL/parser-param assembly).
### Changes
- **`pipeline/parser_page_cap.go`** (new):
- `BuildParserPageCapOverride(parserConfig, dsl, docType, capPages int,
parserComponentName string, familyOf)` — injects the
`ParserConfig[cpnID][family]["pages"]` cap through the same
`override_params` channel production uses. The cap value and family
resolution are injected by the caller, so the function carries no debug
semantics and is reusable for any page-cap scenario.
- `ExtractParserCpnID(dsl, parserComponentName)` — shared Parser cpnID
discovery from (optionally enveloped) DSL.
- `UnwrapCanvasDSL(raw []byte)` — exported single source of truth for
stripping the `{"dsl": {...}}` canvas envelope.
- `pipeline` does **not** import `component` (no reverse dependency);
callers inject `component.ComponentNameParser` /
`component.ParserFileFamily`.
- **`task/pipeline_executor.go`**: removed `injectDebugPageCap` (the
`debugPageCapPages = 2` constant stays in the task package). The debug
branch now calls `pipeline.BuildParserPageCapOverride(...)`.
- **`task/pipeline_executor.go` `warnUnknownComponentParams`**: fixed a
production no-op bug — it passed the enveloped DSL straight to
`ExtractAllComponentParams`, which silently errored and disabled the
unknown-cpnID guard. It now unwraps the envelope first.
- **`task/debug_result_dsl.go`**: reuses `pipeline.UnwrapCanvasDSL`
instead of a third inline envelope-unwrap copy.
### Behavior
No external debug-preview behavior changes. The three original
invariants are preserved exactly:
1. explicit `pages` caps under `cpnID+family` are respected (not
overwritten),
2. an empty family (unknown docType) is a no-op,
3. the injected shape is `[]any{[]any{1, capPages}}` (the
`[]any`-of-`[]any` form `NormalizePDFPages` requires).
## Test plan
- New `pipeline/parser_page_cap_test.go`: `BuildParserPageCapOverride`
(inject / respect-existing / unknown-family no-op / no-Parser no-op),
`ExtractParserCpnID` (enveloped + raw), `UnwrapCanvasDSL`.
- `task/debug_test.go`: `TestInjectDebugPageCap` migrated to the new
helper; new
`TestWarnUnknownComponentParamsDetectsUnknownCPNFromEnvelope` captures
the warning via `zaptest/observer` to prove the envelope no-op bug is
fixed.
- Both `internal/ingestion/pipeline` and `internal/ingestion/task` pass
`build.sh --test` (unit tier).
## Notes
- `TOKEN_CHUNKER_HANDOFF.md` is an unrelated untracked file and was
deliberately **not** included in this PR.
2026-08-06 15:50:39 +08:00
// by pipeline.BuildParserPageCapOverride via Run's override_params channel
// (see TestInjectDebugPageCap).
Feat(ingestion): add canvas pipeline debug (dry-run) mode with View result log (#17538)
## Summary
Adds a side-effect-free DataFlow canvas **debug (dry-run) mode** plus a
**debug run log with a "View result" panel**, so a canvas can be
executed synchronously and inspected end-to-end (per-component progress
and parsed chunks) without persisting anything.
### Dry-run execution (inline parsed chunks)
- `task/debug.go`: `NewDebugTaskContext` builds an in-memory
`TaskContext` with **`KB.ID == ""` — the single debug signal used across
the ingestion pipeline**. A canvas debug run has no knowledgebase, and
production ingestion always supplies one, so `kb_id == ""` occurs ONLY
in debug mode. Components gate their own side effects on this signal
without any dedicated debug vocabulary (the former `CANVAS_DEBUG_DOC_ID`
marker constant is removed).
- `task/pipeline_executor.go`: `validateTaskContext` no longer requires
a KB when `KB.ID == ""` (debug); debug runs return `collectDebugOutput`
(chunks) instead of a no-op; uploaded bytes are delivered as
`inputs['binary']` for doc-less runs; `injectDebugPageCap` caps the
parser to the first pages for a fast preview via the production
`override_params` channel (Parser cpnID + family).
- `component/tokenizer.go`: `shouldHaveEmbedding` skips embedding when
`kb_id == ""` — the embedder is configured on the knowledgebase, so a
debug run has nothing to resolve against and stays side-effect free.
- `chunker/register.go`: chunk images are uploaded to MinIO only when a
KB is present (persist run). **In debug mode the raw image bytes are
intentionally dropped (`delete(ck, "image")`)** — the debug preview does
not render chunk images, and dropping the bytes keeps them out of memory
and out of the Redis-stored debug log. This is a deliberate trade-off,
not an oversight.
- `component/file.go`: pass through in-memory binary bytes, skipping
`doc_id` -> storage resolution.
- `handler/agent.go` + `agent_webhook.go`: detect `dataflow_canvas` and
run a sync debug returning chunks inline on the existing
chat/completions endpoint; reject DataFlow canvases from webhooks (fixes
the previously dead `== "DataFlow"` check; mirrors Python
`agent_api.py`).
- `parser_dispatch.go`: export `ParserFileFamily` for the executor's
page-cap injection.
### Debug run log + "View result"
Mirrors Python's debug-log contract so the front-end can replay each
component's progress and parsed output:
- `task/debug_log_sink.go`: a `DebugLogSink` records every component's
lifecycle into a `[{component_id, trace}]` array (each trace entry
carries `message`, `progress`, `timestamp`, `elapsed_time`). `Flush`
appends a terminal `END` marker whose first trace message is non-empty
so the front-end detects completion. On failure the END marker is
prefixed `[ERROR]` yet still carries the run, so the failure timeline
renders instead of being stuck empty. Timestamps and `elapsed_time` are
in seconds (matching the rest of the app).
- `task/debug_result_dsl.go`: `BuildDebugResultDSL` builds the `dsl` the
END marker carries — the Go analogue of Python's `Graph.__str__` +
END-marker `dsl` in `rag/flow/pipeline.py`. It combines the static DSL
structure (component_name / downstream / params / graph.nodes) with the
run output map (`output["state"][<id>]`) to emit, per component,
`obj.params.outputs[<format>].value` (chunks / text / json / html /
markdown) — the exact keys the front-end `dataflow-result` page reads to
render each step's parsed chunks. Raw embedding vectors (including the
dimension-scoped `q_<dim>_vec` keys) are stripped so the stored log
stays Python-scale.
- `task/pipeline_executor.go`: after the run, attach the built `dsl` to
the END marker via the `ResultSink` capability.
- `handler/agent.go`: `runCanvasPipelineDebug` generates a stable
`message_id` up-front and always flushes the log (success or failure);
`respondWithDebugResult` returns `message_id` in **both** the success
and the error envelope so the front-end can poll the log. The debug-log
endpoint `GET /agents/:id/logs/:message_id` serves the array.
- `web/src/pages/agent/hooks/use-run-dataflow.ts`: on a run failure,
also surface `message_id` via `setMessageId` so the log sheet renders
the failure timeline (the `[ERROR]` END marker is already written).
Guarded by `if (msgId)`, so it is a safe no-op when the back-end does
not return an id.
## Behavioral notes
- Debug parses only the first pages (`debugPageCapPages`) for a fast
preview; an explicit `pages` cap already present in the ParserConfig is
respected.
- Debug mode does not keep chunk images (see above) and does not compute
embeddings — it exercises parse + chunk only.
## Test plan
- Go: `debug_test.go`, `debug_log_sink_test.go` (trace pairing, END
marker, `[ERROR]` prefix, fractional-second timestamp/elapsed_time, size
caps, and a real-pipeline test asserting the END-marker `dsl` carries
non-empty per-component `params.outputs` with chunks),
`debug_result_dsl_test.go` (flat and real nested `output["state"]`
shapes, vector stripping, format priority),
`debug_pages_integration_test.go`, `pipeline_executor_persist_test.go`,
`handler/agent_pipeline_debug_test.go`, `handler/agent_logs_test.go`
(incl. `TestRunCanvasPipelineDebug_ErrorStillExposesMessageID` /
`TestRespondWithDebugResult_ErrorCarriesMessageID` locking `message_id`
on failure), plus updates to `agent_test.go` / `agent_webhook_test.go` /
`chunker/image_upload_test.go` / `tokenizer*.go`.
- `go build ./...` and `./build.sh --test` for affected packages.
🤖 Generated with [CodeBuddy Code](https://cnb.cool/codebuddy)
---------
Co-authored-by: CodeBuddy Code <noreply@tencent.com>
2026-07-31 13:10:08 +08:00
func TestNewDebugTaskContext_InjectsDebugID ( t * testing . T ) {
taskCtx := NewDebugTaskContext ( "t1" , "canvas-1" , "doc.pdf" , [ ] byte ( "page one\fpage two\fpage three" ) )
if taskCtx . Doc . ID == "" {
t . Errorf ( "Doc.ID = %q, want non-empty (uuid)" , taskCtx . Doc . ID )
}
if taskCtx . Doc . ParserConfig != nil {
t . Errorf ( "Doc.ParserConfig = %v, want nil (debug page cap is injected via override_params, not a flat ParserConfig key)" , taskCtx . Doc . ParserConfig )
}
if taskCtx . Doc . KbID != "" {
t . Errorf ( "Doc.KbID = %q, want empty (debug has no KB)" , taskCtx . Doc . KbID )
}
if taskCtx . KB . ID != "" {
t . Errorf ( "KB.ID = %q, want empty (debug has no KB)" , taskCtx . KB . ID )
}
if taskCtx . Tenant . ID != "t1" {
t . Errorf ( "Tenant.ID = %q, want t1" , taskCtx . Tenant . ID )
}
if taskCtx . PipelineID != "canvas-1" {
t . Errorf ( "PipelineID = %q, want canvas-1" , taskCtx . PipelineID )
}
}
// TestExecute_DebugViaEntry proves the entry-point constructor produces a
// valid debug TaskContext that routes through PipelineExecutor.Execute and
// returns the pipeline's chunks WITHOUT persisting (no index insert, no
// pipeline log).
func TestExecute_DebugViaEntry ( t * testing . T ) {
taskCtx := NewDebugTaskContext ( "t1" , "canvas-1" , "doc.pdf" , [ ] byte ( "page one\fpage two\fpage three" ) )
logCalled := false
insertCalled := false
exec , err := NewPipelineExecutor ( taskCtx , "canvas-1" , 0 )
if err != nil {
t . Fatalf ( "NewPipelineExecutor: %v" , err )
}
exec .
WithLoadDSLFunc ( func ( ctx context . Context , canvasID string ) ( string , string , error ) {
return "dsl" , "canvas-1" , nil
} ) .
WithRunPipelineFunc ( func ( ctx context . Context , dsl string ) ( map [ string ] any , string , error ) {
return map [ string ] any {
"chunks" : [ ] map [ string ] any {
{ "text" : "c1" } ,
{ "text" : "c2" } ,
{ "text" : "c3" } ,
{ "text" : "c4" } ,
} ,
} , dsl , nil
} ) .
WithLogCreateFunc ( func ( ctx context . Context , db * gorm . DB , log * entity . PipelineOperationLog ) error {
logCalled = true
return nil
} ) .
WithInsertFunc ( func ( ctx context . Context , chunks [ ] map [ string ] any , baseName , datasetID string ) ( [ ] string , error ) {
insertCalled = true
return nil , nil
} )
result , err := exec . Execute ( context . Background ( ) )
if err != nil {
t . Fatalf ( "Execute: %v" , err )
}
if result == nil {
t . Fatal ( "result is nil" )
}
if len ( result . Chunks ) != 4 {
t . Errorf ( "len(result.Chunks) = %d, want 4" , len ( result . Chunks ) )
}
if logCalled {
t . Error ( "pipeline log should NOT be created in a debug (kb_id == \"\") run" )
}
if insertCalled {
t . Error ( "chunk insert should NOT be called in a debug (kb_id == \"\") run" )
}
}
// TestInjectDebugPageCap verifies the canvas-debug page cap is delivered
// through Run's override_params channel (the existing ParserConfig shape),
// NOT through pipeline inputs. The cap must land at
// ParserConfig[cpnID][family]["pages"], expressed as the JSON-decoded
// []any{[]any{1, N}} form (a list of [from,to] pairs) — cpnID is the Parser
// component's instance id from the DSL and family is the document's filetype
// family. This is the exact shape NormalizeParserConfigPages produces after a
// storage JSON round-trip and the shape the deepdoc pdf parser consumes
// (NormalizePDFPages requires []any, not a Go [][]int).
//
refactor(task): sink parser page-cap override into pipeline package (#17905)
## Summary
Moves the canvas-debug parser page-cap injection out of the `task`
orchestrator and into a **debug-agnostic** `pipeline` helper, so
`PipelineExecutor` keeps only the orchestration skeleton (resolving one
of the P1 review findings: the executor was overloaded with
DSL/parser-param assembly).
### Changes
- **`pipeline/parser_page_cap.go`** (new):
- `BuildParserPageCapOverride(parserConfig, dsl, docType, capPages int,
parserComponentName string, familyOf)` — injects the
`ParserConfig[cpnID][family]["pages"]` cap through the same
`override_params` channel production uses. The cap value and family
resolution are injected by the caller, so the function carries no debug
semantics and is reusable for any page-cap scenario.
- `ExtractParserCpnID(dsl, parserComponentName)` — shared Parser cpnID
discovery from (optionally enveloped) DSL.
- `UnwrapCanvasDSL(raw []byte)` — exported single source of truth for
stripping the `{"dsl": {...}}` canvas envelope.
- `pipeline` does **not** import `component` (no reverse dependency);
callers inject `component.ComponentNameParser` /
`component.ParserFileFamily`.
- **`task/pipeline_executor.go`**: removed `injectDebugPageCap` (the
`debugPageCapPages = 2` constant stays in the task package). The debug
branch now calls `pipeline.BuildParserPageCapOverride(...)`.
- **`task/pipeline_executor.go` `warnUnknownComponentParams`**: fixed a
production no-op bug — it passed the enveloped DSL straight to
`ExtractAllComponentParams`, which silently errored and disabled the
unknown-cpnID guard. It now unwraps the envelope first.
- **`task/debug_result_dsl.go`**: reuses `pipeline.UnwrapCanvasDSL`
instead of a third inline envelope-unwrap copy.
### Behavior
No external debug-preview behavior changes. The three original
invariants are preserved exactly:
1. explicit `pages` caps under `cpnID+family` are respected (not
overwritten),
2. an empty family (unknown docType) is a no-op,
3. the injected shape is `[]any{[]any{1, capPages}}` (the
`[]any`-of-`[]any` form `NormalizePDFPages` requires).
## Test plan
- New `pipeline/parser_page_cap_test.go`: `BuildParserPageCapOverride`
(inject / respect-existing / unknown-family no-op / no-Parser no-op),
`ExtractParserCpnID` (enveloped + raw), `UnwrapCanvasDSL`.
- `task/debug_test.go`: `TestInjectDebugPageCap` migrated to the new
helper; new
`TestWarnUnknownComponentParamsDetectsUnknownCPNFromEnvelope` captures
the warning via `zaptest/observer` to prove the envelope no-op bug is
fixed.
- Both `internal/ingestion/pipeline` and `internal/ingestion/task` pass
`build.sh --test` (unit tier).
## Notes
- `TOKEN_CHUNKER_HANDOFF.md` is an unrelated untracked file and was
deliberately **not** included in this PR.
2026-08-06 15:50:39 +08:00
// The DSL/parser-family knowledge now lives in pipeline.BuildParserPageCapOverride
// (debug-agnostic); this test drives that helper directly to pin the behavior
// the executor relies on.
//
Feat(ingestion): add canvas pipeline debug (dry-run) mode with View result log (#17538)
## Summary
Adds a side-effect-free DataFlow canvas **debug (dry-run) mode** plus a
**debug run log with a "View result" panel**, so a canvas can be
executed synchronously and inspected end-to-end (per-component progress
and parsed chunks) without persisting anything.
### Dry-run execution (inline parsed chunks)
- `task/debug.go`: `NewDebugTaskContext` builds an in-memory
`TaskContext` with **`KB.ID == ""` — the single debug signal used across
the ingestion pipeline**. A canvas debug run has no knowledgebase, and
production ingestion always supplies one, so `kb_id == ""` occurs ONLY
in debug mode. Components gate their own side effects on this signal
without any dedicated debug vocabulary (the former `CANVAS_DEBUG_DOC_ID`
marker constant is removed).
- `task/pipeline_executor.go`: `validateTaskContext` no longer requires
a KB when `KB.ID == ""` (debug); debug runs return `collectDebugOutput`
(chunks) instead of a no-op; uploaded bytes are delivered as
`inputs['binary']` for doc-less runs; `injectDebugPageCap` caps the
parser to the first pages for a fast preview via the production
`override_params` channel (Parser cpnID + family).
- `component/tokenizer.go`: `shouldHaveEmbedding` skips embedding when
`kb_id == ""` — the embedder is configured on the knowledgebase, so a
debug run has nothing to resolve against and stays side-effect free.
- `chunker/register.go`: chunk images are uploaded to MinIO only when a
KB is present (persist run). **In debug mode the raw image bytes are
intentionally dropped (`delete(ck, "image")`)** — the debug preview does
not render chunk images, and dropping the bytes keeps them out of memory
and out of the Redis-stored debug log. This is a deliberate trade-off,
not an oversight.
- `component/file.go`: pass through in-memory binary bytes, skipping
`doc_id` -> storage resolution.
- `handler/agent.go` + `agent_webhook.go`: detect `dataflow_canvas` and
run a sync debug returning chunks inline on the existing
chat/completions endpoint; reject DataFlow canvases from webhooks (fixes
the previously dead `== "DataFlow"` check; mirrors Python
`agent_api.py`).
- `parser_dispatch.go`: export `ParserFileFamily` for the executor's
page-cap injection.
### Debug run log + "View result"
Mirrors Python's debug-log contract so the front-end can replay each
component's progress and parsed output:
- `task/debug_log_sink.go`: a `DebugLogSink` records every component's
lifecycle into a `[{component_id, trace}]` array (each trace entry
carries `message`, `progress`, `timestamp`, `elapsed_time`). `Flush`
appends a terminal `END` marker whose first trace message is non-empty
so the front-end detects completion. On failure the END marker is
prefixed `[ERROR]` yet still carries the run, so the failure timeline
renders instead of being stuck empty. Timestamps and `elapsed_time` are
in seconds (matching the rest of the app).
- `task/debug_result_dsl.go`: `BuildDebugResultDSL` builds the `dsl` the
END marker carries — the Go analogue of Python's `Graph.__str__` +
END-marker `dsl` in `rag/flow/pipeline.py`. It combines the static DSL
structure (component_name / downstream / params / graph.nodes) with the
run output map (`output["state"][<id>]`) to emit, per component,
`obj.params.outputs[<format>].value` (chunks / text / json / html /
markdown) — the exact keys the front-end `dataflow-result` page reads to
render each step's parsed chunks. Raw embedding vectors (including the
dimension-scoped `q_<dim>_vec` keys) are stripped so the stored log
stays Python-scale.
- `task/pipeline_executor.go`: after the run, attach the built `dsl` to
the END marker via the `ResultSink` capability.
- `handler/agent.go`: `runCanvasPipelineDebug` generates a stable
`message_id` up-front and always flushes the log (success or failure);
`respondWithDebugResult` returns `message_id` in **both** the success
and the error envelope so the front-end can poll the log. The debug-log
endpoint `GET /agents/:id/logs/:message_id` serves the array.
- `web/src/pages/agent/hooks/use-run-dataflow.ts`: on a run failure,
also surface `message_id` via `setMessageId` so the log sheet renders
the failure timeline (the `[ERROR]` END marker is already written).
Guarded by `if (msgId)`, so it is a safe no-op when the back-end does
not return an id.
## Behavioral notes
- Debug parses only the first pages (`debugPageCapPages`) for a fast
preview; an explicit `pages` cap already present in the ParserConfig is
respected.
- Debug mode does not keep chunk images (see above) and does not compute
embeddings — it exercises parse + chunk only.
## Test plan
- Go: `debug_test.go`, `debug_log_sink_test.go` (trace pairing, END
marker, `[ERROR]` prefix, fractional-second timestamp/elapsed_time, size
caps, and a real-pipeline test asserting the END-marker `dsl` carries
non-empty per-component `params.outputs` with chunks),
`debug_result_dsl_test.go` (flat and real nested `output["state"]`
shapes, vector stripping, format priority),
`debug_pages_integration_test.go`, `pipeline_executor_persist_test.go`,
`handler/agent_pipeline_debug_test.go`, `handler/agent_logs_test.go`
(incl. `TestRunCanvasPipelineDebug_ErrorStillExposesMessageID` /
`TestRespondWithDebugResult_ErrorCarriesMessageID` locking `message_id`
on failure), plus updates to `agent_test.go` / `agent_webhook_test.go` /
`chunker/image_upload_test.go` / `tokenizer*.go`.
- `go build ./...` and `./build.sh --test` for affected packages.
🤖 Generated with [CodeBuddy Code](https://cnb.cool/codebuddy)
---------
Co-authored-by: CodeBuddy Code <noreply@tencent.com>
2026-07-31 13:10:08 +08:00
// It also pins the regression: a flat top-level "pages" key would be dropped
// by the override_params merge and never reach the parser, so the cap must be
// nested under the cpnID.
func TestInjectDebugPageCap ( t * testing . T ) {
templatePath := filepath . Join ( taskRepoRoot ( t ) , "internal" , "ingestion" , "pipeline" , "template" , "ingestion_pipeline_general.json" )
raw , err := os . ReadFile ( templatePath )
if err != nil {
t . Fatalf ( "read template: %v" , err )
}
var envelope struct {
DSL json . RawMessage ` json:"dsl" `
}
if err := json . Unmarshal ( raw , & envelope ) ; err != nil {
t . Fatalf ( "unmarshal template envelope: %v" , err )
}
dsl := string ( envelope . DSL )
// Discover the Parser cpnID the same way the executor does.
schemas , err := pipeline . ExtractAllComponentParams ( envelope . DSL )
if err != nil {
t . Fatalf ( "ExtractAllComponentParams: %v" , err )
}
var parserCpnID string
for _ , s := range schemas {
if s . ComponentName == component . ComponentNameParser {
parserCpnID = s . CpnID
}
}
if parserCpnID == "" {
t . Fatal ( "template has no Parser component" )
}
refactor(task): sink parser page-cap override into pipeline package (#17905)
## Summary
Moves the canvas-debug parser page-cap injection out of the `task`
orchestrator and into a **debug-agnostic** `pipeline` helper, so
`PipelineExecutor` keeps only the orchestration skeleton (resolving one
of the P1 review findings: the executor was overloaded with
DSL/parser-param assembly).
### Changes
- **`pipeline/parser_page_cap.go`** (new):
- `BuildParserPageCapOverride(parserConfig, dsl, docType, capPages int,
parserComponentName string, familyOf)` — injects the
`ParserConfig[cpnID][family]["pages"]` cap through the same
`override_params` channel production uses. The cap value and family
resolution are injected by the caller, so the function carries no debug
semantics and is reusable for any page-cap scenario.
- `ExtractParserCpnID(dsl, parserComponentName)` — shared Parser cpnID
discovery from (optionally enveloped) DSL.
- `UnwrapCanvasDSL(raw []byte)` — exported single source of truth for
stripping the `{"dsl": {...}}` canvas envelope.
- `pipeline` does **not** import `component` (no reverse dependency);
callers inject `component.ComponentNameParser` /
`component.ParserFileFamily`.
- **`task/pipeline_executor.go`**: removed `injectDebugPageCap` (the
`debugPageCapPages = 2` constant stays in the task package). The debug
branch now calls `pipeline.BuildParserPageCapOverride(...)`.
- **`task/pipeline_executor.go` `warnUnknownComponentParams`**: fixed a
production no-op bug — it passed the enveloped DSL straight to
`ExtractAllComponentParams`, which silently errored and disabled the
unknown-cpnID guard. It now unwraps the envelope first.
- **`task/debug_result_dsl.go`**: reuses `pipeline.UnwrapCanvasDSL`
instead of a third inline envelope-unwrap copy.
### Behavior
No external debug-preview behavior changes. The three original
invariants are preserved exactly:
1. explicit `pages` caps under `cpnID+family` are respected (not
overwritten),
2. an empty family (unknown docType) is a no-op,
3. the injected shape is `[]any{[]any{1, capPages}}` (the
`[]any`-of-`[]any` form `NormalizePDFPages` requires).
## Test plan
- New `pipeline/parser_page_cap_test.go`: `BuildParserPageCapOverride`
(inject / respect-existing / unknown-family no-op / no-Parser no-op),
`ExtractParserCpnID` (enveloped + raw), `UnwrapCanvasDSL`.
- `task/debug_test.go`: `TestInjectDebugPageCap` migrated to the new
helper; new
`TestWarnUnknownComponentParamsDetectsUnknownCPNFromEnvelope` captures
the warning via `zaptest/observer` to prove the envelope no-op bug is
fixed.
- Both `internal/ingestion/pipeline` and `internal/ingestion/task` pass
`build.sh --test` (unit tier).
## Notes
- `TOKEN_CHUNKER_HANDOFF.md` is an unrelated untracked file and was
deliberately **not** included in this PR.
2026-08-06 15:50:39 +08:00
apply := func ( cfg map [ string ] any , dslArg string , docType string ) map [ string ] any {
return pipeline . BuildParserPageCapOverride (
cfg , [ ] byte ( dslArg ) , docType , debugPageCapPages ,
component . ComponentNameParser , component . ParserFileFamily )
}
Feat(ingestion): add canvas pipeline debug (dry-run) mode with View result log (#17538)
## Summary
Adds a side-effect-free DataFlow canvas **debug (dry-run) mode** plus a
**debug run log with a "View result" panel**, so a canvas can be
executed synchronously and inspected end-to-end (per-component progress
and parsed chunks) without persisting anything.
### Dry-run execution (inline parsed chunks)
- `task/debug.go`: `NewDebugTaskContext` builds an in-memory
`TaskContext` with **`KB.ID == ""` — the single debug signal used across
the ingestion pipeline**. A canvas debug run has no knowledgebase, and
production ingestion always supplies one, so `kb_id == ""` occurs ONLY
in debug mode. Components gate their own side effects on this signal
without any dedicated debug vocabulary (the former `CANVAS_DEBUG_DOC_ID`
marker constant is removed).
- `task/pipeline_executor.go`: `validateTaskContext` no longer requires
a KB when `KB.ID == ""` (debug); debug runs return `collectDebugOutput`
(chunks) instead of a no-op; uploaded bytes are delivered as
`inputs['binary']` for doc-less runs; `injectDebugPageCap` caps the
parser to the first pages for a fast preview via the production
`override_params` channel (Parser cpnID + family).
- `component/tokenizer.go`: `shouldHaveEmbedding` skips embedding when
`kb_id == ""` — the embedder is configured on the knowledgebase, so a
debug run has nothing to resolve against and stays side-effect free.
- `chunker/register.go`: chunk images are uploaded to MinIO only when a
KB is present (persist run). **In debug mode the raw image bytes are
intentionally dropped (`delete(ck, "image")`)** — the debug preview does
not render chunk images, and dropping the bytes keeps them out of memory
and out of the Redis-stored debug log. This is a deliberate trade-off,
not an oversight.
- `component/file.go`: pass through in-memory binary bytes, skipping
`doc_id` -> storage resolution.
- `handler/agent.go` + `agent_webhook.go`: detect `dataflow_canvas` and
run a sync debug returning chunks inline on the existing
chat/completions endpoint; reject DataFlow canvases from webhooks (fixes
the previously dead `== "DataFlow"` check; mirrors Python
`agent_api.py`).
- `parser_dispatch.go`: export `ParserFileFamily` for the executor's
page-cap injection.
### Debug run log + "View result"
Mirrors Python's debug-log contract so the front-end can replay each
component's progress and parsed output:
- `task/debug_log_sink.go`: a `DebugLogSink` records every component's
lifecycle into a `[{component_id, trace}]` array (each trace entry
carries `message`, `progress`, `timestamp`, `elapsed_time`). `Flush`
appends a terminal `END` marker whose first trace message is non-empty
so the front-end detects completion. On failure the END marker is
prefixed `[ERROR]` yet still carries the run, so the failure timeline
renders instead of being stuck empty. Timestamps and `elapsed_time` are
in seconds (matching the rest of the app).
- `task/debug_result_dsl.go`: `BuildDebugResultDSL` builds the `dsl` the
END marker carries — the Go analogue of Python's `Graph.__str__` +
END-marker `dsl` in `rag/flow/pipeline.py`. It combines the static DSL
structure (component_name / downstream / params / graph.nodes) with the
run output map (`output["state"][<id>]`) to emit, per component,
`obj.params.outputs[<format>].value` (chunks / text / json / html /
markdown) — the exact keys the front-end `dataflow-result` page reads to
render each step's parsed chunks. Raw embedding vectors (including the
dimension-scoped `q_<dim>_vec` keys) are stripped so the stored log
stays Python-scale.
- `task/pipeline_executor.go`: after the run, attach the built `dsl` to
the END marker via the `ResultSink` capability.
- `handler/agent.go`: `runCanvasPipelineDebug` generates a stable
`message_id` up-front and always flushes the log (success or failure);
`respondWithDebugResult` returns `message_id` in **both** the success
and the error envelope so the front-end can poll the log. The debug-log
endpoint `GET /agents/:id/logs/:message_id` serves the array.
- `web/src/pages/agent/hooks/use-run-dataflow.ts`: on a run failure,
also surface `message_id` via `setMessageId` so the log sheet renders
the failure timeline (the `[ERROR]` END marker is already written).
Guarded by `if (msgId)`, so it is a safe no-op when the back-end does
not return an id.
## Behavioral notes
- Debug parses only the first pages (`debugPageCapPages`) for a fast
preview; an explicit `pages` cap already present in the ParserConfig is
respected.
- Debug mode does not keep chunk images (see above) and does not compute
embeddings — it exercises parse + chunk only.
## Test plan
- Go: `debug_test.go`, `debug_log_sink_test.go` (trace pairing, END
marker, `[ERROR]` prefix, fractional-second timestamp/elapsed_time, size
caps, and a real-pipeline test asserting the END-marker `dsl` carries
non-empty per-component `params.outputs` with chunks),
`debug_result_dsl_test.go` (flat and real nested `output["state"]`
shapes, vector stripping, format priority),
`debug_pages_integration_test.go`, `pipeline_executor_persist_test.go`,
`handler/agent_pipeline_debug_test.go`, `handler/agent_logs_test.go`
(incl. `TestRunCanvasPipelineDebug_ErrorStillExposesMessageID` /
`TestRespondWithDebugResult_ErrorCarriesMessageID` locking `message_id`
on failure), plus updates to `agent_test.go` / `agent_webhook_test.go` /
`chunker/image_upload_test.go` / `tokenizer*.go`.
- `go build ./...` and `./build.sh --test` for affected packages.
🤖 Generated with [CodeBuddy Code](https://cnb.cool/codebuddy)
---------
Co-authored-by: CodeBuddy Code <noreply@tencent.com>
2026-07-31 13:10:08 +08:00
t . Run ( "pdf injects [1,2] under cpnID+family" , func ( t * testing . T ) {
parserConfig := map [ string ] any { }
refactor(task): sink parser page-cap override into pipeline package (#17905)
## Summary
Moves the canvas-debug parser page-cap injection out of the `task`
orchestrator and into a **debug-agnostic** `pipeline` helper, so
`PipelineExecutor` keeps only the orchestration skeleton (resolving one
of the P1 review findings: the executor was overloaded with
DSL/parser-param assembly).
### Changes
- **`pipeline/parser_page_cap.go`** (new):
- `BuildParserPageCapOverride(parserConfig, dsl, docType, capPages int,
parserComponentName string, familyOf)` — injects the
`ParserConfig[cpnID][family]["pages"]` cap through the same
`override_params` channel production uses. The cap value and family
resolution are injected by the caller, so the function carries no debug
semantics and is reusable for any page-cap scenario.
- `ExtractParserCpnID(dsl, parserComponentName)` — shared Parser cpnID
discovery from (optionally enveloped) DSL.
- `UnwrapCanvasDSL(raw []byte)` — exported single source of truth for
stripping the `{"dsl": {...}}` canvas envelope.
- `pipeline` does **not** import `component` (no reverse dependency);
callers inject `component.ComponentNameParser` /
`component.ParserFileFamily`.
- **`task/pipeline_executor.go`**: removed `injectDebugPageCap` (the
`debugPageCapPages = 2` constant stays in the task package). The debug
branch now calls `pipeline.BuildParserPageCapOverride(...)`.
- **`task/pipeline_executor.go` `warnUnknownComponentParams`**: fixed a
production no-op bug — it passed the enveloped DSL straight to
`ExtractAllComponentParams`, which silently errored and disabled the
unknown-cpnID guard. It now unwraps the envelope first.
- **`task/debug_result_dsl.go`**: reuses `pipeline.UnwrapCanvasDSL`
instead of a third inline envelope-unwrap copy.
### Behavior
No external debug-preview behavior changes. The three original
invariants are preserved exactly:
1. explicit `pages` caps under `cpnID+family` are respected (not
overwritten),
2. an empty family (unknown docType) is a no-op,
3. the injected shape is `[]any{[]any{1, capPages}}` (the
`[]any`-of-`[]any` form `NormalizePDFPages` requires).
## Test plan
- New `pipeline/parser_page_cap_test.go`: `BuildParserPageCapOverride`
(inject / respect-existing / unknown-family no-op / no-Parser no-op),
`ExtractParserCpnID` (enveloped + raw), `UnwrapCanvasDSL`.
- `task/debug_test.go`: `TestInjectDebugPageCap` migrated to the new
helper; new
`TestWarnUnknownComponentParamsDetectsUnknownCPNFromEnvelope` captures
the warning via `zaptest/observer` to prove the envelope no-op bug is
fixed.
- Both `internal/ingestion/pipeline` and `internal/ingestion/task` pass
`build.sh --test` (unit tier).
## Notes
- `TOKEN_CHUNKER_HANDOFF.md` is an unrelated untracked file and was
deliberately **not** included in this PR.
2026-08-06 15:50:39 +08:00
apply ( parserConfig , dsl , "pdf" )
Feat(ingestion): add canvas pipeline debug (dry-run) mode with View result log (#17538)
## Summary
Adds a side-effect-free DataFlow canvas **debug (dry-run) mode** plus a
**debug run log with a "View result" panel**, so a canvas can be
executed synchronously and inspected end-to-end (per-component progress
and parsed chunks) without persisting anything.
### Dry-run execution (inline parsed chunks)
- `task/debug.go`: `NewDebugTaskContext` builds an in-memory
`TaskContext` with **`KB.ID == ""` — the single debug signal used across
the ingestion pipeline**. A canvas debug run has no knowledgebase, and
production ingestion always supplies one, so `kb_id == ""` occurs ONLY
in debug mode. Components gate their own side effects on this signal
without any dedicated debug vocabulary (the former `CANVAS_DEBUG_DOC_ID`
marker constant is removed).
- `task/pipeline_executor.go`: `validateTaskContext` no longer requires
a KB when `KB.ID == ""` (debug); debug runs return `collectDebugOutput`
(chunks) instead of a no-op; uploaded bytes are delivered as
`inputs['binary']` for doc-less runs; `injectDebugPageCap` caps the
parser to the first pages for a fast preview via the production
`override_params` channel (Parser cpnID + family).
- `component/tokenizer.go`: `shouldHaveEmbedding` skips embedding when
`kb_id == ""` — the embedder is configured on the knowledgebase, so a
debug run has nothing to resolve against and stays side-effect free.
- `chunker/register.go`: chunk images are uploaded to MinIO only when a
KB is present (persist run). **In debug mode the raw image bytes are
intentionally dropped (`delete(ck, "image")`)** — the debug preview does
not render chunk images, and dropping the bytes keeps them out of memory
and out of the Redis-stored debug log. This is a deliberate trade-off,
not an oversight.
- `component/file.go`: pass through in-memory binary bytes, skipping
`doc_id` -> storage resolution.
- `handler/agent.go` + `agent_webhook.go`: detect `dataflow_canvas` and
run a sync debug returning chunks inline on the existing
chat/completions endpoint; reject DataFlow canvases from webhooks (fixes
the previously dead `== "DataFlow"` check; mirrors Python
`agent_api.py`).
- `parser_dispatch.go`: export `ParserFileFamily` for the executor's
page-cap injection.
### Debug run log + "View result"
Mirrors Python's debug-log contract so the front-end can replay each
component's progress and parsed output:
- `task/debug_log_sink.go`: a `DebugLogSink` records every component's
lifecycle into a `[{component_id, trace}]` array (each trace entry
carries `message`, `progress`, `timestamp`, `elapsed_time`). `Flush`
appends a terminal `END` marker whose first trace message is non-empty
so the front-end detects completion. On failure the END marker is
prefixed `[ERROR]` yet still carries the run, so the failure timeline
renders instead of being stuck empty. Timestamps and `elapsed_time` are
in seconds (matching the rest of the app).
- `task/debug_result_dsl.go`: `BuildDebugResultDSL` builds the `dsl` the
END marker carries — the Go analogue of Python's `Graph.__str__` +
END-marker `dsl` in `rag/flow/pipeline.py`. It combines the static DSL
structure (component_name / downstream / params / graph.nodes) with the
run output map (`output["state"][<id>]`) to emit, per component,
`obj.params.outputs[<format>].value` (chunks / text / json / html /
markdown) — the exact keys the front-end `dataflow-result` page reads to
render each step's parsed chunks. Raw embedding vectors (including the
dimension-scoped `q_<dim>_vec` keys) are stripped so the stored log
stays Python-scale.
- `task/pipeline_executor.go`: after the run, attach the built `dsl` to
the END marker via the `ResultSink` capability.
- `handler/agent.go`: `runCanvasPipelineDebug` generates a stable
`message_id` up-front and always flushes the log (success or failure);
`respondWithDebugResult` returns `message_id` in **both** the success
and the error envelope so the front-end can poll the log. The debug-log
endpoint `GET /agents/:id/logs/:message_id` serves the array.
- `web/src/pages/agent/hooks/use-run-dataflow.ts`: on a run failure,
also surface `message_id` via `setMessageId` so the log sheet renders
the failure timeline (the `[ERROR]` END marker is already written).
Guarded by `if (msgId)`, so it is a safe no-op when the back-end does
not return an id.
## Behavioral notes
- Debug parses only the first pages (`debugPageCapPages`) for a fast
preview; an explicit `pages` cap already present in the ParserConfig is
respected.
- Debug mode does not keep chunk images (see above) and does not compute
embeddings — it exercises parse + chunk only.
## Test plan
- Go: `debug_test.go`, `debug_log_sink_test.go` (trace pairing, END
marker, `[ERROR]` prefix, fractional-second timestamp/elapsed_time, size
caps, and a real-pipeline test asserting the END-marker `dsl` carries
non-empty per-component `params.outputs` with chunks),
`debug_result_dsl_test.go` (flat and real nested `output["state"]`
shapes, vector stripping, format priority),
`debug_pages_integration_test.go`, `pipeline_executor_persist_test.go`,
`handler/agent_pipeline_debug_test.go`, `handler/agent_logs_test.go`
(incl. `TestRunCanvasPipelineDebug_ErrorStillExposesMessageID` /
`TestRespondWithDebugResult_ErrorCarriesMessageID` locking `message_id`
on failure), plus updates to `agent_test.go` / `agent_webhook_test.go` /
`chunker/image_upload_test.go` / `tokenizer*.go`.
- `go build ./...` and `./build.sh --test` for affected packages.
🤖 Generated with [CodeBuddy Code](https://cnb.cool/codebuddy)
---------
Co-authored-by: CodeBuddy Code <noreply@tencent.com>
2026-07-31 13:10:08 +08:00
famEntry , ok := parserConfig [ parserCpnID ] . ( map [ string ] any )
if ! ok {
t . Fatalf ( "parserConfig[%q] = %T, want map[string]any" , parserCpnID , parserConfig [ parserCpnID ] )
}
pdf , ok := famEntry [ "pdf" ] . ( map [ string ] any )
if ! ok {
t . Fatalf ( "parserConfig[%q][\"pdf\"] = %T, want map[string]any" , parserCpnID , famEntry [ "pdf" ] )
}
if ! reflect . DeepEqual ( pdf [ "pages" ] , [ ] any { [ ] any { 1 , debugPageCapPages } } ) {
t . Errorf ( "parserConfig[%q][\"pdf\"][\"pages\"] = %v, want [[1, %d]]" , parserCpnID , pdf [ "pages" ] , debugPageCapPages )
}
} )
t . Run ( "docx injects under docx family" , func ( t * testing . T ) {
parserConfig := map [ string ] any { }
refactor(task): sink parser page-cap override into pipeline package (#17905)
## Summary
Moves the canvas-debug parser page-cap injection out of the `task`
orchestrator and into a **debug-agnostic** `pipeline` helper, so
`PipelineExecutor` keeps only the orchestration skeleton (resolving one
of the P1 review findings: the executor was overloaded with
DSL/parser-param assembly).
### Changes
- **`pipeline/parser_page_cap.go`** (new):
- `BuildParserPageCapOverride(parserConfig, dsl, docType, capPages int,
parserComponentName string, familyOf)` — injects the
`ParserConfig[cpnID][family]["pages"]` cap through the same
`override_params` channel production uses. The cap value and family
resolution are injected by the caller, so the function carries no debug
semantics and is reusable for any page-cap scenario.
- `ExtractParserCpnID(dsl, parserComponentName)` — shared Parser cpnID
discovery from (optionally enveloped) DSL.
- `UnwrapCanvasDSL(raw []byte)` — exported single source of truth for
stripping the `{"dsl": {...}}` canvas envelope.
- `pipeline` does **not** import `component` (no reverse dependency);
callers inject `component.ComponentNameParser` /
`component.ParserFileFamily`.
- **`task/pipeline_executor.go`**: removed `injectDebugPageCap` (the
`debugPageCapPages = 2` constant stays in the task package). The debug
branch now calls `pipeline.BuildParserPageCapOverride(...)`.
- **`task/pipeline_executor.go` `warnUnknownComponentParams`**: fixed a
production no-op bug — it passed the enveloped DSL straight to
`ExtractAllComponentParams`, which silently errored and disabled the
unknown-cpnID guard. It now unwraps the envelope first.
- **`task/debug_result_dsl.go`**: reuses `pipeline.UnwrapCanvasDSL`
instead of a third inline envelope-unwrap copy.
### Behavior
No external debug-preview behavior changes. The three original
invariants are preserved exactly:
1. explicit `pages` caps under `cpnID+family` are respected (not
overwritten),
2. an empty family (unknown docType) is a no-op,
3. the injected shape is `[]any{[]any{1, capPages}}` (the
`[]any`-of-`[]any` form `NormalizePDFPages` requires).
## Test plan
- New `pipeline/parser_page_cap_test.go`: `BuildParserPageCapOverride`
(inject / respect-existing / unknown-family no-op / no-Parser no-op),
`ExtractParserCpnID` (enveloped + raw), `UnwrapCanvasDSL`.
- `task/debug_test.go`: `TestInjectDebugPageCap` migrated to the new
helper; new
`TestWarnUnknownComponentParamsDetectsUnknownCPNFromEnvelope` captures
the warning via `zaptest/observer` to prove the envelope no-op bug is
fixed.
- Both `internal/ingestion/pipeline` and `internal/ingestion/task` pass
`build.sh --test` (unit tier).
## Notes
- `TOKEN_CHUNKER_HANDOFF.md` is an unrelated untracked file and was
deliberately **not** included in this PR.
2026-08-06 15:50:39 +08:00
apply ( parserConfig , dsl , "docx" )
Feat(ingestion): add canvas pipeline debug (dry-run) mode with View result log (#17538)
## Summary
Adds a side-effect-free DataFlow canvas **debug (dry-run) mode** plus a
**debug run log with a "View result" panel**, so a canvas can be
executed synchronously and inspected end-to-end (per-component progress
and parsed chunks) without persisting anything.
### Dry-run execution (inline parsed chunks)
- `task/debug.go`: `NewDebugTaskContext` builds an in-memory
`TaskContext` with **`KB.ID == ""` — the single debug signal used across
the ingestion pipeline**. A canvas debug run has no knowledgebase, and
production ingestion always supplies one, so `kb_id == ""` occurs ONLY
in debug mode. Components gate their own side effects on this signal
without any dedicated debug vocabulary (the former `CANVAS_DEBUG_DOC_ID`
marker constant is removed).
- `task/pipeline_executor.go`: `validateTaskContext` no longer requires
a KB when `KB.ID == ""` (debug); debug runs return `collectDebugOutput`
(chunks) instead of a no-op; uploaded bytes are delivered as
`inputs['binary']` for doc-less runs; `injectDebugPageCap` caps the
parser to the first pages for a fast preview via the production
`override_params` channel (Parser cpnID + family).
- `component/tokenizer.go`: `shouldHaveEmbedding` skips embedding when
`kb_id == ""` — the embedder is configured on the knowledgebase, so a
debug run has nothing to resolve against and stays side-effect free.
- `chunker/register.go`: chunk images are uploaded to MinIO only when a
KB is present (persist run). **In debug mode the raw image bytes are
intentionally dropped (`delete(ck, "image")`)** — the debug preview does
not render chunk images, and dropping the bytes keeps them out of memory
and out of the Redis-stored debug log. This is a deliberate trade-off,
not an oversight.
- `component/file.go`: pass through in-memory binary bytes, skipping
`doc_id` -> storage resolution.
- `handler/agent.go` + `agent_webhook.go`: detect `dataflow_canvas` and
run a sync debug returning chunks inline on the existing
chat/completions endpoint; reject DataFlow canvases from webhooks (fixes
the previously dead `== "DataFlow"` check; mirrors Python
`agent_api.py`).
- `parser_dispatch.go`: export `ParserFileFamily` for the executor's
page-cap injection.
### Debug run log + "View result"
Mirrors Python's debug-log contract so the front-end can replay each
component's progress and parsed output:
- `task/debug_log_sink.go`: a `DebugLogSink` records every component's
lifecycle into a `[{component_id, trace}]` array (each trace entry
carries `message`, `progress`, `timestamp`, `elapsed_time`). `Flush`
appends a terminal `END` marker whose first trace message is non-empty
so the front-end detects completion. On failure the END marker is
prefixed `[ERROR]` yet still carries the run, so the failure timeline
renders instead of being stuck empty. Timestamps and `elapsed_time` are
in seconds (matching the rest of the app).
- `task/debug_result_dsl.go`: `BuildDebugResultDSL` builds the `dsl` the
END marker carries — the Go analogue of Python's `Graph.__str__` +
END-marker `dsl` in `rag/flow/pipeline.py`. It combines the static DSL
structure (component_name / downstream / params / graph.nodes) with the
run output map (`output["state"][<id>]`) to emit, per component,
`obj.params.outputs[<format>].value` (chunks / text / json / html /
markdown) — the exact keys the front-end `dataflow-result` page reads to
render each step's parsed chunks. Raw embedding vectors (including the
dimension-scoped `q_<dim>_vec` keys) are stripped so the stored log
stays Python-scale.
- `task/pipeline_executor.go`: after the run, attach the built `dsl` to
the END marker via the `ResultSink` capability.
- `handler/agent.go`: `runCanvasPipelineDebug` generates a stable
`message_id` up-front and always flushes the log (success or failure);
`respondWithDebugResult` returns `message_id` in **both** the success
and the error envelope so the front-end can poll the log. The debug-log
endpoint `GET /agents/:id/logs/:message_id` serves the array.
- `web/src/pages/agent/hooks/use-run-dataflow.ts`: on a run failure,
also surface `message_id` via `setMessageId` so the log sheet renders
the failure timeline (the `[ERROR]` END marker is already written).
Guarded by `if (msgId)`, so it is a safe no-op when the back-end does
not return an id.
## Behavioral notes
- Debug parses only the first pages (`debugPageCapPages`) for a fast
preview; an explicit `pages` cap already present in the ParserConfig is
respected.
- Debug mode does not keep chunk images (see above) and does not compute
embeddings — it exercises parse + chunk only.
## Test plan
- Go: `debug_test.go`, `debug_log_sink_test.go` (trace pairing, END
marker, `[ERROR]` prefix, fractional-second timestamp/elapsed_time, size
caps, and a real-pipeline test asserting the END-marker `dsl` carries
non-empty per-component `params.outputs` with chunks),
`debug_result_dsl_test.go` (flat and real nested `output["state"]`
shapes, vector stripping, format priority),
`debug_pages_integration_test.go`, `pipeline_executor_persist_test.go`,
`handler/agent_pipeline_debug_test.go`, `handler/agent_logs_test.go`
(incl. `TestRunCanvasPipelineDebug_ErrorStillExposesMessageID` /
`TestRespondWithDebugResult_ErrorCarriesMessageID` locking `message_id`
on failure), plus updates to `agent_test.go` / `agent_webhook_test.go` /
`chunker/image_upload_test.go` / `tokenizer*.go`.
- `go build ./...` and `./build.sh --test` for affected packages.
🤖 Generated with [CodeBuddy Code](https://cnb.cool/codebuddy)
---------
Co-authored-by: CodeBuddy Code <noreply@tencent.com>
2026-07-31 13:10:08 +08:00
famEntry , ok := parserConfig [ parserCpnID ] . ( map [ string ] any )
if ! ok {
t . Fatalf ( "parserConfig[%q] = %T, want map[string]any" , parserCpnID , parserConfig [ parserCpnID ] )
}
docx , ok := famEntry [ "docx" ] . ( map [ string ] any )
if ! ok {
t . Fatalf ( "parserConfig[%q][\"docx\"] = %T, want map[string]any" , parserCpnID , famEntry [ "docx" ] )
}
if ! reflect . DeepEqual ( docx [ "pages" ] , [ ] any { [ ] any { 1 , debugPageCapPages } } ) {
t . Errorf ( "parserConfig[%q][\"docx\"][\"pages\"] = %v, want [[1, %d]]" , parserCpnID , docx [ "pages" ] , debugPageCapPages )
}
} )
t . Run ( "empty docType is a no-op" , func ( t * testing . T ) {
parserConfig := map [ string ] any { }
refactor(task): sink parser page-cap override into pipeline package (#17905)
## Summary
Moves the canvas-debug parser page-cap injection out of the `task`
orchestrator and into a **debug-agnostic** `pipeline` helper, so
`PipelineExecutor` keeps only the orchestration skeleton (resolving one
of the P1 review findings: the executor was overloaded with
DSL/parser-param assembly).
### Changes
- **`pipeline/parser_page_cap.go`** (new):
- `BuildParserPageCapOverride(parserConfig, dsl, docType, capPages int,
parserComponentName string, familyOf)` — injects the
`ParserConfig[cpnID][family]["pages"]` cap through the same
`override_params` channel production uses. The cap value and family
resolution are injected by the caller, so the function carries no debug
semantics and is reusable for any page-cap scenario.
- `ExtractParserCpnID(dsl, parserComponentName)` — shared Parser cpnID
discovery from (optionally enveloped) DSL.
- `UnwrapCanvasDSL(raw []byte)` — exported single source of truth for
stripping the `{"dsl": {...}}` canvas envelope.
- `pipeline` does **not** import `component` (no reverse dependency);
callers inject `component.ComponentNameParser` /
`component.ParserFileFamily`.
- **`task/pipeline_executor.go`**: removed `injectDebugPageCap` (the
`debugPageCapPages = 2` constant stays in the task package). The debug
branch now calls `pipeline.BuildParserPageCapOverride(...)`.
- **`task/pipeline_executor.go` `warnUnknownComponentParams`**: fixed a
production no-op bug — it passed the enveloped DSL straight to
`ExtractAllComponentParams`, which silently errored and disabled the
unknown-cpnID guard. It now unwraps the envelope first.
- **`task/debug_result_dsl.go`**: reuses `pipeline.UnwrapCanvasDSL`
instead of a third inline envelope-unwrap copy.
### Behavior
No external debug-preview behavior changes. The three original
invariants are preserved exactly:
1. explicit `pages` caps under `cpnID+family` are respected (not
overwritten),
2. an empty family (unknown docType) is a no-op,
3. the injected shape is `[]any{[]any{1, capPages}}` (the
`[]any`-of-`[]any` form `NormalizePDFPages` requires).
## Test plan
- New `pipeline/parser_page_cap_test.go`: `BuildParserPageCapOverride`
(inject / respect-existing / unknown-family no-op / no-Parser no-op),
`ExtractParserCpnID` (enveloped + raw), `UnwrapCanvasDSL`.
- `task/debug_test.go`: `TestInjectDebugPageCap` migrated to the new
helper; new
`TestWarnUnknownComponentParamsDetectsUnknownCPNFromEnvelope` captures
the warning via `zaptest/observer` to prove the envelope no-op bug is
fixed.
- Both `internal/ingestion/pipeline` and `internal/ingestion/task` pass
`build.sh --test` (unit tier).
## Notes
- `TOKEN_CHUNKER_HANDOFF.md` is an unrelated untracked file and was
deliberately **not** included in this PR.
2026-08-06 15:50:39 +08:00
apply ( parserConfig , dsl , "" )
Feat(ingestion): add canvas pipeline debug (dry-run) mode with View result log (#17538)
## Summary
Adds a side-effect-free DataFlow canvas **debug (dry-run) mode** plus a
**debug run log with a "View result" panel**, so a canvas can be
executed synchronously and inspected end-to-end (per-component progress
and parsed chunks) without persisting anything.
### Dry-run execution (inline parsed chunks)
- `task/debug.go`: `NewDebugTaskContext` builds an in-memory
`TaskContext` with **`KB.ID == ""` — the single debug signal used across
the ingestion pipeline**. A canvas debug run has no knowledgebase, and
production ingestion always supplies one, so `kb_id == ""` occurs ONLY
in debug mode. Components gate their own side effects on this signal
without any dedicated debug vocabulary (the former `CANVAS_DEBUG_DOC_ID`
marker constant is removed).
- `task/pipeline_executor.go`: `validateTaskContext` no longer requires
a KB when `KB.ID == ""` (debug); debug runs return `collectDebugOutput`
(chunks) instead of a no-op; uploaded bytes are delivered as
`inputs['binary']` for doc-less runs; `injectDebugPageCap` caps the
parser to the first pages for a fast preview via the production
`override_params` channel (Parser cpnID + family).
- `component/tokenizer.go`: `shouldHaveEmbedding` skips embedding when
`kb_id == ""` — the embedder is configured on the knowledgebase, so a
debug run has nothing to resolve against and stays side-effect free.
- `chunker/register.go`: chunk images are uploaded to MinIO only when a
KB is present (persist run). **In debug mode the raw image bytes are
intentionally dropped (`delete(ck, "image")`)** — the debug preview does
not render chunk images, and dropping the bytes keeps them out of memory
and out of the Redis-stored debug log. This is a deliberate trade-off,
not an oversight.
- `component/file.go`: pass through in-memory binary bytes, skipping
`doc_id` -> storage resolution.
- `handler/agent.go` + `agent_webhook.go`: detect `dataflow_canvas` and
run a sync debug returning chunks inline on the existing
chat/completions endpoint; reject DataFlow canvases from webhooks (fixes
the previously dead `== "DataFlow"` check; mirrors Python
`agent_api.py`).
- `parser_dispatch.go`: export `ParserFileFamily` for the executor's
page-cap injection.
### Debug run log + "View result"
Mirrors Python's debug-log contract so the front-end can replay each
component's progress and parsed output:
- `task/debug_log_sink.go`: a `DebugLogSink` records every component's
lifecycle into a `[{component_id, trace}]` array (each trace entry
carries `message`, `progress`, `timestamp`, `elapsed_time`). `Flush`
appends a terminal `END` marker whose first trace message is non-empty
so the front-end detects completion. On failure the END marker is
prefixed `[ERROR]` yet still carries the run, so the failure timeline
renders instead of being stuck empty. Timestamps and `elapsed_time` are
in seconds (matching the rest of the app).
- `task/debug_result_dsl.go`: `BuildDebugResultDSL` builds the `dsl` the
END marker carries — the Go analogue of Python's `Graph.__str__` +
END-marker `dsl` in `rag/flow/pipeline.py`. It combines the static DSL
structure (component_name / downstream / params / graph.nodes) with the
run output map (`output["state"][<id>]`) to emit, per component,
`obj.params.outputs[<format>].value` (chunks / text / json / html /
markdown) — the exact keys the front-end `dataflow-result` page reads to
render each step's parsed chunks. Raw embedding vectors (including the
dimension-scoped `q_<dim>_vec` keys) are stripped so the stored log
stays Python-scale.
- `task/pipeline_executor.go`: after the run, attach the built `dsl` to
the END marker via the `ResultSink` capability.
- `handler/agent.go`: `runCanvasPipelineDebug` generates a stable
`message_id` up-front and always flushes the log (success or failure);
`respondWithDebugResult` returns `message_id` in **both** the success
and the error envelope so the front-end can poll the log. The debug-log
endpoint `GET /agents/:id/logs/:message_id` serves the array.
- `web/src/pages/agent/hooks/use-run-dataflow.ts`: on a run failure,
also surface `message_id` via `setMessageId` so the log sheet renders
the failure timeline (the `[ERROR]` END marker is already written).
Guarded by `if (msgId)`, so it is a safe no-op when the back-end does
not return an id.
## Behavioral notes
- Debug parses only the first pages (`debugPageCapPages`) for a fast
preview; an explicit `pages` cap already present in the ParserConfig is
respected.
- Debug mode does not keep chunk images (see above) and does not compute
embeddings — it exercises parse + chunk only.
## Test plan
- Go: `debug_test.go`, `debug_log_sink_test.go` (trace pairing, END
marker, `[ERROR]` prefix, fractional-second timestamp/elapsed_time, size
caps, and a real-pipeline test asserting the END-marker `dsl` carries
non-empty per-component `params.outputs` with chunks),
`debug_result_dsl_test.go` (flat and real nested `output["state"]`
shapes, vector stripping, format priority),
`debug_pages_integration_test.go`, `pipeline_executor_persist_test.go`,
`handler/agent_pipeline_debug_test.go`, `handler/agent_logs_test.go`
(incl. `TestRunCanvasPipelineDebug_ErrorStillExposesMessageID` /
`TestRespondWithDebugResult_ErrorCarriesMessageID` locking `message_id`
on failure), plus updates to `agent_test.go` / `agent_webhook_test.go` /
`chunker/image_upload_test.go` / `tokenizer*.go`.
- `go build ./...` and `./build.sh --test` for affected packages.
🤖 Generated with [CodeBuddy Code](https://cnb.cool/codebuddy)
---------
Co-authored-by: CodeBuddy Code <noreply@tencent.com>
2026-07-31 13:10:08 +08:00
if len ( parserConfig ) != 0 {
t . Errorf ( "parserConfig = %v, want empty (no family derivable from empty docType)" , parserConfig )
}
} )
t . Run ( "does not clobber existing parser params" , func ( t * testing . T ) {
parserConfig := map [ string ] any {
parserCpnID : map [ string ] any {
"pdf" : map [ string ] any { "parse_method" : "deepdoc" } ,
} ,
}
refactor(task): sink parser page-cap override into pipeline package (#17905)
## Summary
Moves the canvas-debug parser page-cap injection out of the `task`
orchestrator and into a **debug-agnostic** `pipeline` helper, so
`PipelineExecutor` keeps only the orchestration skeleton (resolving one
of the P1 review findings: the executor was overloaded with
DSL/parser-param assembly).
### Changes
- **`pipeline/parser_page_cap.go`** (new):
- `BuildParserPageCapOverride(parserConfig, dsl, docType, capPages int,
parserComponentName string, familyOf)` — injects the
`ParserConfig[cpnID][family]["pages"]` cap through the same
`override_params` channel production uses. The cap value and family
resolution are injected by the caller, so the function carries no debug
semantics and is reusable for any page-cap scenario.
- `ExtractParserCpnID(dsl, parserComponentName)` — shared Parser cpnID
discovery from (optionally enveloped) DSL.
- `UnwrapCanvasDSL(raw []byte)` — exported single source of truth for
stripping the `{"dsl": {...}}` canvas envelope.
- `pipeline` does **not** import `component` (no reverse dependency);
callers inject `component.ComponentNameParser` /
`component.ParserFileFamily`.
- **`task/pipeline_executor.go`**: removed `injectDebugPageCap` (the
`debugPageCapPages = 2` constant stays in the task package). The debug
branch now calls `pipeline.BuildParserPageCapOverride(...)`.
- **`task/pipeline_executor.go` `warnUnknownComponentParams`**: fixed a
production no-op bug — it passed the enveloped DSL straight to
`ExtractAllComponentParams`, which silently errored and disabled the
unknown-cpnID guard. It now unwraps the envelope first.
- **`task/debug_result_dsl.go`**: reuses `pipeline.UnwrapCanvasDSL`
instead of a third inline envelope-unwrap copy.
### Behavior
No external debug-preview behavior changes. The three original
invariants are preserved exactly:
1. explicit `pages` caps under `cpnID+family` are respected (not
overwritten),
2. an empty family (unknown docType) is a no-op,
3. the injected shape is `[]any{[]any{1, capPages}}` (the
`[]any`-of-`[]any` form `NormalizePDFPages` requires).
## Test plan
- New `pipeline/parser_page_cap_test.go`: `BuildParserPageCapOverride`
(inject / respect-existing / unknown-family no-op / no-Parser no-op),
`ExtractParserCpnID` (enveloped + raw), `UnwrapCanvasDSL`.
- `task/debug_test.go`: `TestInjectDebugPageCap` migrated to the new
helper; new
`TestWarnUnknownComponentParamsDetectsUnknownCPNFromEnvelope` captures
the warning via `zaptest/observer` to prove the envelope no-op bug is
fixed.
- Both `internal/ingestion/pipeline` and `internal/ingestion/task` pass
`build.sh --test` (unit tier).
## Notes
- `TOKEN_CHUNKER_HANDOFF.md` is an unrelated untracked file and was
deliberately **not** included in this PR.
2026-08-06 15:50:39 +08:00
apply ( parserConfig , dsl , "pdf" )
Feat(ingestion): add canvas pipeline debug (dry-run) mode with View result log (#17538)
## Summary
Adds a side-effect-free DataFlow canvas **debug (dry-run) mode** plus a
**debug run log with a "View result" panel**, so a canvas can be
executed synchronously and inspected end-to-end (per-component progress
and parsed chunks) without persisting anything.
### Dry-run execution (inline parsed chunks)
- `task/debug.go`: `NewDebugTaskContext` builds an in-memory
`TaskContext` with **`KB.ID == ""` — the single debug signal used across
the ingestion pipeline**. A canvas debug run has no knowledgebase, and
production ingestion always supplies one, so `kb_id == ""` occurs ONLY
in debug mode. Components gate their own side effects on this signal
without any dedicated debug vocabulary (the former `CANVAS_DEBUG_DOC_ID`
marker constant is removed).
- `task/pipeline_executor.go`: `validateTaskContext` no longer requires
a KB when `KB.ID == ""` (debug); debug runs return `collectDebugOutput`
(chunks) instead of a no-op; uploaded bytes are delivered as
`inputs['binary']` for doc-less runs; `injectDebugPageCap` caps the
parser to the first pages for a fast preview via the production
`override_params` channel (Parser cpnID + family).
- `component/tokenizer.go`: `shouldHaveEmbedding` skips embedding when
`kb_id == ""` — the embedder is configured on the knowledgebase, so a
debug run has nothing to resolve against and stays side-effect free.
- `chunker/register.go`: chunk images are uploaded to MinIO only when a
KB is present (persist run). **In debug mode the raw image bytes are
intentionally dropped (`delete(ck, "image")`)** — the debug preview does
not render chunk images, and dropping the bytes keeps them out of memory
and out of the Redis-stored debug log. This is a deliberate trade-off,
not an oversight.
- `component/file.go`: pass through in-memory binary bytes, skipping
`doc_id` -> storage resolution.
- `handler/agent.go` + `agent_webhook.go`: detect `dataflow_canvas` and
run a sync debug returning chunks inline on the existing
chat/completions endpoint; reject DataFlow canvases from webhooks (fixes
the previously dead `== "DataFlow"` check; mirrors Python
`agent_api.py`).
- `parser_dispatch.go`: export `ParserFileFamily` for the executor's
page-cap injection.
### Debug run log + "View result"
Mirrors Python's debug-log contract so the front-end can replay each
component's progress and parsed output:
- `task/debug_log_sink.go`: a `DebugLogSink` records every component's
lifecycle into a `[{component_id, trace}]` array (each trace entry
carries `message`, `progress`, `timestamp`, `elapsed_time`). `Flush`
appends a terminal `END` marker whose first trace message is non-empty
so the front-end detects completion. On failure the END marker is
prefixed `[ERROR]` yet still carries the run, so the failure timeline
renders instead of being stuck empty. Timestamps and `elapsed_time` are
in seconds (matching the rest of the app).
- `task/debug_result_dsl.go`: `BuildDebugResultDSL` builds the `dsl` the
END marker carries — the Go analogue of Python's `Graph.__str__` +
END-marker `dsl` in `rag/flow/pipeline.py`. It combines the static DSL
structure (component_name / downstream / params / graph.nodes) with the
run output map (`output["state"][<id>]`) to emit, per component,
`obj.params.outputs[<format>].value` (chunks / text / json / html /
markdown) — the exact keys the front-end `dataflow-result` page reads to
render each step's parsed chunks. Raw embedding vectors (including the
dimension-scoped `q_<dim>_vec` keys) are stripped so the stored log
stays Python-scale.
- `task/pipeline_executor.go`: after the run, attach the built `dsl` to
the END marker via the `ResultSink` capability.
- `handler/agent.go`: `runCanvasPipelineDebug` generates a stable
`message_id` up-front and always flushes the log (success or failure);
`respondWithDebugResult` returns `message_id` in **both** the success
and the error envelope so the front-end can poll the log. The debug-log
endpoint `GET /agents/:id/logs/:message_id` serves the array.
- `web/src/pages/agent/hooks/use-run-dataflow.ts`: on a run failure,
also surface `message_id` via `setMessageId` so the log sheet renders
the failure timeline (the `[ERROR]` END marker is already written).
Guarded by `if (msgId)`, so it is a safe no-op when the back-end does
not return an id.
## Behavioral notes
- Debug parses only the first pages (`debugPageCapPages`) for a fast
preview; an explicit `pages` cap already present in the ParserConfig is
respected.
- Debug mode does not keep chunk images (see above) and does not compute
embeddings — it exercises parse + chunk only.
## Test plan
- Go: `debug_test.go`, `debug_log_sink_test.go` (trace pairing, END
marker, `[ERROR]` prefix, fractional-second timestamp/elapsed_time, size
caps, and a real-pipeline test asserting the END-marker `dsl` carries
non-empty per-component `params.outputs` with chunks),
`debug_result_dsl_test.go` (flat and real nested `output["state"]`
shapes, vector stripping, format priority),
`debug_pages_integration_test.go`, `pipeline_executor_persist_test.go`,
`handler/agent_pipeline_debug_test.go`, `handler/agent_logs_test.go`
(incl. `TestRunCanvasPipelineDebug_ErrorStillExposesMessageID` /
`TestRespondWithDebugResult_ErrorCarriesMessageID` locking `message_id`
on failure), plus updates to `agent_test.go` / `agent_webhook_test.go` /
`chunker/image_upload_test.go` / `tokenizer*.go`.
- `go build ./...` and `./build.sh --test` for affected packages.
🤖 Generated with [CodeBuddy Code](https://cnb.cool/codebuddy)
---------
Co-authored-by: CodeBuddy Code <noreply@tencent.com>
2026-07-31 13:10:08 +08:00
famEntry := parserConfig [ parserCpnID ] . ( map [ string ] any )
pdf := famEntry [ "pdf" ] . ( map [ string ] any )
if pdf [ "parse_method" ] != "deepdoc" {
t . Errorf ( "parserConfig[%q][\"pdf\"][\"parse_method\"] = %v, want deepdoc (existing params must be preserved)" , parserCpnID , pdf [ "parse_method" ] )
}
if ! reflect . DeepEqual ( pdf [ "pages" ] , [ ] any { [ ] any { 1 , debugPageCapPages } } ) {
t . Errorf ( "parserConfig[%q][\"pdf\"][\"pages\"] = %v, want [[1, %d]]" , parserCpnID , pdf [ "pages" ] , debugPageCapPages )
}
} )
t . Run ( "envelope dsl form (production shape)" , func ( t * testing . T ) {
// In production dsl is the canvas envelope {"dsl": {"components": ...}},
refactor(task): sink parser page-cap override into pipeline package (#17905)
## Summary
Moves the canvas-debug parser page-cap injection out of the `task`
orchestrator and into a **debug-agnostic** `pipeline` helper, so
`PipelineExecutor` keeps only the orchestration skeleton (resolving one
of the P1 review findings: the executor was overloaded with
DSL/parser-param assembly).
### Changes
- **`pipeline/parser_page_cap.go`** (new):
- `BuildParserPageCapOverride(parserConfig, dsl, docType, capPages int,
parserComponentName string, familyOf)` — injects the
`ParserConfig[cpnID][family]["pages"]` cap through the same
`override_params` channel production uses. The cap value and family
resolution are injected by the caller, so the function carries no debug
semantics and is reusable for any page-cap scenario.
- `ExtractParserCpnID(dsl, parserComponentName)` — shared Parser cpnID
discovery from (optionally enveloped) DSL.
- `UnwrapCanvasDSL(raw []byte)` — exported single source of truth for
stripping the `{"dsl": {...}}` canvas envelope.
- `pipeline` does **not** import `component` (no reverse dependency);
callers inject `component.ComponentNameParser` /
`component.ParserFileFamily`.
- **`task/pipeline_executor.go`**: removed `injectDebugPageCap` (the
`debugPageCapPages = 2` constant stays in the task package). The debug
branch now calls `pipeline.BuildParserPageCapOverride(...)`.
- **`task/pipeline_executor.go` `warnUnknownComponentParams`**: fixed a
production no-op bug — it passed the enveloped DSL straight to
`ExtractAllComponentParams`, which silently errored and disabled the
unknown-cpnID guard. It now unwraps the envelope first.
- **`task/debug_result_dsl.go`**: reuses `pipeline.UnwrapCanvasDSL`
instead of a third inline envelope-unwrap copy.
### Behavior
No external debug-preview behavior changes. The three original
invariants are preserved exactly:
1. explicit `pages` caps under `cpnID+family` are respected (not
overwritten),
2. an empty family (unknown docType) is a no-op,
3. the injected shape is `[]any{[]any{1, capPages}}` (the
`[]any`-of-`[]any` form `NormalizePDFPages` requires).
## Test plan
- New `pipeline/parser_page_cap_test.go`: `BuildParserPageCapOverride`
(inject / respect-existing / unknown-family no-op / no-Parser no-op),
`ExtractParserCpnID` (enveloped + raw), `UnwrapCanvasDSL`.
- `task/debug_test.go`: `TestInjectDebugPageCap` migrated to the new
helper; new
`TestWarnUnknownComponentParamsDetectsUnknownCPNFromEnvelope` captures
the warning via `zaptest/observer` to prove the envelope no-op bug is
fixed.
- Both `internal/ingestion/pipeline` and `internal/ingestion/task` pass
`build.sh --test` (unit tier).
## Notes
- `TOKEN_CHUNKER_HANDOFF.md` is an unrelated untracked file and was
deliberately **not** included in this PR.
2026-08-06 15:50:39 +08:00
// not the bare components map. The helper must still find the Parser
// cpnID after unwrapping.
Feat(ingestion): add canvas pipeline debug (dry-run) mode with View result log (#17538)
## Summary
Adds a side-effect-free DataFlow canvas **debug (dry-run) mode** plus a
**debug run log with a "View result" panel**, so a canvas can be
executed synchronously and inspected end-to-end (per-component progress
and parsed chunks) without persisting anything.
### Dry-run execution (inline parsed chunks)
- `task/debug.go`: `NewDebugTaskContext` builds an in-memory
`TaskContext` with **`KB.ID == ""` — the single debug signal used across
the ingestion pipeline**. A canvas debug run has no knowledgebase, and
production ingestion always supplies one, so `kb_id == ""` occurs ONLY
in debug mode. Components gate their own side effects on this signal
without any dedicated debug vocabulary (the former `CANVAS_DEBUG_DOC_ID`
marker constant is removed).
- `task/pipeline_executor.go`: `validateTaskContext` no longer requires
a KB when `KB.ID == ""` (debug); debug runs return `collectDebugOutput`
(chunks) instead of a no-op; uploaded bytes are delivered as
`inputs['binary']` for doc-less runs; `injectDebugPageCap` caps the
parser to the first pages for a fast preview via the production
`override_params` channel (Parser cpnID + family).
- `component/tokenizer.go`: `shouldHaveEmbedding` skips embedding when
`kb_id == ""` — the embedder is configured on the knowledgebase, so a
debug run has nothing to resolve against and stays side-effect free.
- `chunker/register.go`: chunk images are uploaded to MinIO only when a
KB is present (persist run). **In debug mode the raw image bytes are
intentionally dropped (`delete(ck, "image")`)** — the debug preview does
not render chunk images, and dropping the bytes keeps them out of memory
and out of the Redis-stored debug log. This is a deliberate trade-off,
not an oversight.
- `component/file.go`: pass through in-memory binary bytes, skipping
`doc_id` -> storage resolution.
- `handler/agent.go` + `agent_webhook.go`: detect `dataflow_canvas` and
run a sync debug returning chunks inline on the existing
chat/completions endpoint; reject DataFlow canvases from webhooks (fixes
the previously dead `== "DataFlow"` check; mirrors Python
`agent_api.py`).
- `parser_dispatch.go`: export `ParserFileFamily` for the executor's
page-cap injection.
### Debug run log + "View result"
Mirrors Python's debug-log contract so the front-end can replay each
component's progress and parsed output:
- `task/debug_log_sink.go`: a `DebugLogSink` records every component's
lifecycle into a `[{component_id, trace}]` array (each trace entry
carries `message`, `progress`, `timestamp`, `elapsed_time`). `Flush`
appends a terminal `END` marker whose first trace message is non-empty
so the front-end detects completion. On failure the END marker is
prefixed `[ERROR]` yet still carries the run, so the failure timeline
renders instead of being stuck empty. Timestamps and `elapsed_time` are
in seconds (matching the rest of the app).
- `task/debug_result_dsl.go`: `BuildDebugResultDSL` builds the `dsl` the
END marker carries — the Go analogue of Python's `Graph.__str__` +
END-marker `dsl` in `rag/flow/pipeline.py`. It combines the static DSL
structure (component_name / downstream / params / graph.nodes) with the
run output map (`output["state"][<id>]`) to emit, per component,
`obj.params.outputs[<format>].value` (chunks / text / json / html /
markdown) — the exact keys the front-end `dataflow-result` page reads to
render each step's parsed chunks. Raw embedding vectors (including the
dimension-scoped `q_<dim>_vec` keys) are stripped so the stored log
stays Python-scale.
- `task/pipeline_executor.go`: after the run, attach the built `dsl` to
the END marker via the `ResultSink` capability.
- `handler/agent.go`: `runCanvasPipelineDebug` generates a stable
`message_id` up-front and always flushes the log (success or failure);
`respondWithDebugResult` returns `message_id` in **both** the success
and the error envelope so the front-end can poll the log. The debug-log
endpoint `GET /agents/:id/logs/:message_id` serves the array.
- `web/src/pages/agent/hooks/use-run-dataflow.ts`: on a run failure,
also surface `message_id` via `setMessageId` so the log sheet renders
the failure timeline (the `[ERROR]` END marker is already written).
Guarded by `if (msgId)`, so it is a safe no-op when the back-end does
not return an id.
## Behavioral notes
- Debug parses only the first pages (`debugPageCapPages`) for a fast
preview; an explicit `pages` cap already present in the ParserConfig is
respected.
- Debug mode does not keep chunk images (see above) and does not compute
embeddings — it exercises parse + chunk only.
## Test plan
- Go: `debug_test.go`, `debug_log_sink_test.go` (trace pairing, END
marker, `[ERROR]` prefix, fractional-second timestamp/elapsed_time, size
caps, and a real-pipeline test asserting the END-marker `dsl` carries
non-empty per-component `params.outputs` with chunks),
`debug_result_dsl_test.go` (flat and real nested `output["state"]`
shapes, vector stripping, format priority),
`debug_pages_integration_test.go`, `pipeline_executor_persist_test.go`,
`handler/agent_pipeline_debug_test.go`, `handler/agent_logs_test.go`
(incl. `TestRunCanvasPipelineDebug_ErrorStillExposesMessageID` /
`TestRespondWithDebugResult_ErrorCarriesMessageID` locking `message_id`
on failure), plus updates to `agent_test.go` / `agent_webhook_test.go` /
`chunker/image_upload_test.go` / `tokenizer*.go`.
- `go build ./...` and `./build.sh --test` for affected packages.
🤖 Generated with [CodeBuddy Code](https://cnb.cool/codebuddy)
---------
Co-authored-by: CodeBuddy Code <noreply@tencent.com>
2026-07-31 13:10:08 +08:00
wrapped := fmt . Sprintf ( ` { "dsl":%s} ` , string ( envelope . DSL ) )
parserConfig := map [ string ] any { }
refactor(task): sink parser page-cap override into pipeline package (#17905)
## Summary
Moves the canvas-debug parser page-cap injection out of the `task`
orchestrator and into a **debug-agnostic** `pipeline` helper, so
`PipelineExecutor` keeps only the orchestration skeleton (resolving one
of the P1 review findings: the executor was overloaded with
DSL/parser-param assembly).
### Changes
- **`pipeline/parser_page_cap.go`** (new):
- `BuildParserPageCapOverride(parserConfig, dsl, docType, capPages int,
parserComponentName string, familyOf)` — injects the
`ParserConfig[cpnID][family]["pages"]` cap through the same
`override_params` channel production uses. The cap value and family
resolution are injected by the caller, so the function carries no debug
semantics and is reusable for any page-cap scenario.
- `ExtractParserCpnID(dsl, parserComponentName)` — shared Parser cpnID
discovery from (optionally enveloped) DSL.
- `UnwrapCanvasDSL(raw []byte)` — exported single source of truth for
stripping the `{"dsl": {...}}` canvas envelope.
- `pipeline` does **not** import `component` (no reverse dependency);
callers inject `component.ComponentNameParser` /
`component.ParserFileFamily`.
- **`task/pipeline_executor.go`**: removed `injectDebugPageCap` (the
`debugPageCapPages = 2` constant stays in the task package). The debug
branch now calls `pipeline.BuildParserPageCapOverride(...)`.
- **`task/pipeline_executor.go` `warnUnknownComponentParams`**: fixed a
production no-op bug — it passed the enveloped DSL straight to
`ExtractAllComponentParams`, which silently errored and disabled the
unknown-cpnID guard. It now unwraps the envelope first.
- **`task/debug_result_dsl.go`**: reuses `pipeline.UnwrapCanvasDSL`
instead of a third inline envelope-unwrap copy.
### Behavior
No external debug-preview behavior changes. The three original
invariants are preserved exactly:
1. explicit `pages` caps under `cpnID+family` are respected (not
overwritten),
2. an empty family (unknown docType) is a no-op,
3. the injected shape is `[]any{[]any{1, capPages}}` (the
`[]any`-of-`[]any` form `NormalizePDFPages` requires).
## Test plan
- New `pipeline/parser_page_cap_test.go`: `BuildParserPageCapOverride`
(inject / respect-existing / unknown-family no-op / no-Parser no-op),
`ExtractParserCpnID` (enveloped + raw), `UnwrapCanvasDSL`.
- `task/debug_test.go`: `TestInjectDebugPageCap` migrated to the new
helper; new
`TestWarnUnknownComponentParamsDetectsUnknownCPNFromEnvelope` captures
the warning via `zaptest/observer` to prove the envelope no-op bug is
fixed.
- Both `internal/ingestion/pipeline` and `internal/ingestion/task` pass
`build.sh --test` (unit tier).
## Notes
- `TOKEN_CHUNKER_HANDOFF.md` is an unrelated untracked file and was
deliberately **not** included in this PR.
2026-08-06 15:50:39 +08:00
apply ( parserConfig , wrapped , "pdf" )
Feat(ingestion): add canvas pipeline debug (dry-run) mode with View result log (#17538)
## Summary
Adds a side-effect-free DataFlow canvas **debug (dry-run) mode** plus a
**debug run log with a "View result" panel**, so a canvas can be
executed synchronously and inspected end-to-end (per-component progress
and parsed chunks) without persisting anything.
### Dry-run execution (inline parsed chunks)
- `task/debug.go`: `NewDebugTaskContext` builds an in-memory
`TaskContext` with **`KB.ID == ""` — the single debug signal used across
the ingestion pipeline**. A canvas debug run has no knowledgebase, and
production ingestion always supplies one, so `kb_id == ""` occurs ONLY
in debug mode. Components gate their own side effects on this signal
without any dedicated debug vocabulary (the former `CANVAS_DEBUG_DOC_ID`
marker constant is removed).
- `task/pipeline_executor.go`: `validateTaskContext` no longer requires
a KB when `KB.ID == ""` (debug); debug runs return `collectDebugOutput`
(chunks) instead of a no-op; uploaded bytes are delivered as
`inputs['binary']` for doc-less runs; `injectDebugPageCap` caps the
parser to the first pages for a fast preview via the production
`override_params` channel (Parser cpnID + family).
- `component/tokenizer.go`: `shouldHaveEmbedding` skips embedding when
`kb_id == ""` — the embedder is configured on the knowledgebase, so a
debug run has nothing to resolve against and stays side-effect free.
- `chunker/register.go`: chunk images are uploaded to MinIO only when a
KB is present (persist run). **In debug mode the raw image bytes are
intentionally dropped (`delete(ck, "image")`)** — the debug preview does
not render chunk images, and dropping the bytes keeps them out of memory
and out of the Redis-stored debug log. This is a deliberate trade-off,
not an oversight.
- `component/file.go`: pass through in-memory binary bytes, skipping
`doc_id` -> storage resolution.
- `handler/agent.go` + `agent_webhook.go`: detect `dataflow_canvas` and
run a sync debug returning chunks inline on the existing
chat/completions endpoint; reject DataFlow canvases from webhooks (fixes
the previously dead `== "DataFlow"` check; mirrors Python
`agent_api.py`).
- `parser_dispatch.go`: export `ParserFileFamily` for the executor's
page-cap injection.
### Debug run log + "View result"
Mirrors Python's debug-log contract so the front-end can replay each
component's progress and parsed output:
- `task/debug_log_sink.go`: a `DebugLogSink` records every component's
lifecycle into a `[{component_id, trace}]` array (each trace entry
carries `message`, `progress`, `timestamp`, `elapsed_time`). `Flush`
appends a terminal `END` marker whose first trace message is non-empty
so the front-end detects completion. On failure the END marker is
prefixed `[ERROR]` yet still carries the run, so the failure timeline
renders instead of being stuck empty. Timestamps and `elapsed_time` are
in seconds (matching the rest of the app).
- `task/debug_result_dsl.go`: `BuildDebugResultDSL` builds the `dsl` the
END marker carries — the Go analogue of Python's `Graph.__str__` +
END-marker `dsl` in `rag/flow/pipeline.py`. It combines the static DSL
structure (component_name / downstream / params / graph.nodes) with the
run output map (`output["state"][<id>]`) to emit, per component,
`obj.params.outputs[<format>].value` (chunks / text / json / html /
markdown) — the exact keys the front-end `dataflow-result` page reads to
render each step's parsed chunks. Raw embedding vectors (including the
dimension-scoped `q_<dim>_vec` keys) are stripped so the stored log
stays Python-scale.
- `task/pipeline_executor.go`: after the run, attach the built `dsl` to
the END marker via the `ResultSink` capability.
- `handler/agent.go`: `runCanvasPipelineDebug` generates a stable
`message_id` up-front and always flushes the log (success or failure);
`respondWithDebugResult` returns `message_id` in **both** the success
and the error envelope so the front-end can poll the log. The debug-log
endpoint `GET /agents/:id/logs/:message_id` serves the array.
- `web/src/pages/agent/hooks/use-run-dataflow.ts`: on a run failure,
also surface `message_id` via `setMessageId` so the log sheet renders
the failure timeline (the `[ERROR]` END marker is already written).
Guarded by `if (msgId)`, so it is a safe no-op when the back-end does
not return an id.
## Behavioral notes
- Debug parses only the first pages (`debugPageCapPages`) for a fast
preview; an explicit `pages` cap already present in the ParserConfig is
respected.
- Debug mode does not keep chunk images (see above) and does not compute
embeddings — it exercises parse + chunk only.
## Test plan
- Go: `debug_test.go`, `debug_log_sink_test.go` (trace pairing, END
marker, `[ERROR]` prefix, fractional-second timestamp/elapsed_time, size
caps, and a real-pipeline test asserting the END-marker `dsl` carries
non-empty per-component `params.outputs` with chunks),
`debug_result_dsl_test.go` (flat and real nested `output["state"]`
shapes, vector stripping, format priority),
`debug_pages_integration_test.go`, `pipeline_executor_persist_test.go`,
`handler/agent_pipeline_debug_test.go`, `handler/agent_logs_test.go`
(incl. `TestRunCanvasPipelineDebug_ErrorStillExposesMessageID` /
`TestRespondWithDebugResult_ErrorCarriesMessageID` locking `message_id`
on failure), plus updates to `agent_test.go` / `agent_webhook_test.go` /
`chunker/image_upload_test.go` / `tokenizer*.go`.
- `go build ./...` and `./build.sh --test` for affected packages.
🤖 Generated with [CodeBuddy Code](https://cnb.cool/codebuddy)
---------
Co-authored-by: CodeBuddy Code <noreply@tencent.com>
2026-07-31 13:10:08 +08:00
famEntry , ok := parserConfig [ parserCpnID ] . ( map [ string ] any )
if ! ok {
t . Fatalf ( "parserConfig[%q] = %T, want map[string]any (envelope dsl must unwrap)" , parserCpnID , parserConfig [ parserCpnID ] )
}
pdf , ok := famEntry [ "pdf" ] . ( map [ string ] any )
if ! ok {
t . Fatalf ( "parserConfig[%q][\"pdf\"] = %T, want map[string]any" , parserCpnID , famEntry [ "pdf" ] )
}
if ! reflect . DeepEqual ( pdf [ "pages" ] , [ ] any { [ ] any { 1 , debugPageCapPages } } ) {
t . Errorf ( "parserConfig[%q][\"pdf\"][\"pages\"] = %v, want [[1, %d]] (envelope dsl not unwrapped?)" , parserCpnID , pdf [ "pages" ] , debugPageCapPages )
}
} )
t . Run ( "respects explicit caller-supplied cap" , func ( t * testing . T ) {
// When the document already carries an explicit cpnID+family page cap,
// the debug default must NOT override it (so a wider/narrower cap wins).
parserConfig := map [ string ] any {
parserCpnID : map [ string ] any {
"pdf" : map [ string ] any { "pages" : [ ] any { [ ] any { 1 , 1000000 } } } ,
} ,
}
refactor(task): sink parser page-cap override into pipeline package (#17905)
## Summary
Moves the canvas-debug parser page-cap injection out of the `task`
orchestrator and into a **debug-agnostic** `pipeline` helper, so
`PipelineExecutor` keeps only the orchestration skeleton (resolving one
of the P1 review findings: the executor was overloaded with
DSL/parser-param assembly).
### Changes
- **`pipeline/parser_page_cap.go`** (new):
- `BuildParserPageCapOverride(parserConfig, dsl, docType, capPages int,
parserComponentName string, familyOf)` — injects the
`ParserConfig[cpnID][family]["pages"]` cap through the same
`override_params` channel production uses. The cap value and family
resolution are injected by the caller, so the function carries no debug
semantics and is reusable for any page-cap scenario.
- `ExtractParserCpnID(dsl, parserComponentName)` — shared Parser cpnID
discovery from (optionally enveloped) DSL.
- `UnwrapCanvasDSL(raw []byte)` — exported single source of truth for
stripping the `{"dsl": {...}}` canvas envelope.
- `pipeline` does **not** import `component` (no reverse dependency);
callers inject `component.ComponentNameParser` /
`component.ParserFileFamily`.
- **`task/pipeline_executor.go`**: removed `injectDebugPageCap` (the
`debugPageCapPages = 2` constant stays in the task package). The debug
branch now calls `pipeline.BuildParserPageCapOverride(...)`.
- **`task/pipeline_executor.go` `warnUnknownComponentParams`**: fixed a
production no-op bug — it passed the enveloped DSL straight to
`ExtractAllComponentParams`, which silently errored and disabled the
unknown-cpnID guard. It now unwraps the envelope first.
- **`task/debug_result_dsl.go`**: reuses `pipeline.UnwrapCanvasDSL`
instead of a third inline envelope-unwrap copy.
### Behavior
No external debug-preview behavior changes. The three original
invariants are preserved exactly:
1. explicit `pages` caps under `cpnID+family` are respected (not
overwritten),
2. an empty family (unknown docType) is a no-op,
3. the injected shape is `[]any{[]any{1, capPages}}` (the
`[]any`-of-`[]any` form `NormalizePDFPages` requires).
## Test plan
- New `pipeline/parser_page_cap_test.go`: `BuildParserPageCapOverride`
(inject / respect-existing / unknown-family no-op / no-Parser no-op),
`ExtractParserCpnID` (enveloped + raw), `UnwrapCanvasDSL`.
- `task/debug_test.go`: `TestInjectDebugPageCap` migrated to the new
helper; new
`TestWarnUnknownComponentParamsDetectsUnknownCPNFromEnvelope` captures
the warning via `zaptest/observer` to prove the envelope no-op bug is
fixed.
- Both `internal/ingestion/pipeline` and `internal/ingestion/task` pass
`build.sh --test` (unit tier).
## Notes
- `TOKEN_CHUNKER_HANDOFF.md` is an unrelated untracked file and was
deliberately **not** included in this PR.
2026-08-06 15:50:39 +08:00
apply ( parserConfig , dsl , "pdf" )
Feat(ingestion): add canvas pipeline debug (dry-run) mode with View result log (#17538)
## Summary
Adds a side-effect-free DataFlow canvas **debug (dry-run) mode** plus a
**debug run log with a "View result" panel**, so a canvas can be
executed synchronously and inspected end-to-end (per-component progress
and parsed chunks) without persisting anything.
### Dry-run execution (inline parsed chunks)
- `task/debug.go`: `NewDebugTaskContext` builds an in-memory
`TaskContext` with **`KB.ID == ""` — the single debug signal used across
the ingestion pipeline**. A canvas debug run has no knowledgebase, and
production ingestion always supplies one, so `kb_id == ""` occurs ONLY
in debug mode. Components gate their own side effects on this signal
without any dedicated debug vocabulary (the former `CANVAS_DEBUG_DOC_ID`
marker constant is removed).
- `task/pipeline_executor.go`: `validateTaskContext` no longer requires
a KB when `KB.ID == ""` (debug); debug runs return `collectDebugOutput`
(chunks) instead of a no-op; uploaded bytes are delivered as
`inputs['binary']` for doc-less runs; `injectDebugPageCap` caps the
parser to the first pages for a fast preview via the production
`override_params` channel (Parser cpnID + family).
- `component/tokenizer.go`: `shouldHaveEmbedding` skips embedding when
`kb_id == ""` — the embedder is configured on the knowledgebase, so a
debug run has nothing to resolve against and stays side-effect free.
- `chunker/register.go`: chunk images are uploaded to MinIO only when a
KB is present (persist run). **In debug mode the raw image bytes are
intentionally dropped (`delete(ck, "image")`)** — the debug preview does
not render chunk images, and dropping the bytes keeps them out of memory
and out of the Redis-stored debug log. This is a deliberate trade-off,
not an oversight.
- `component/file.go`: pass through in-memory binary bytes, skipping
`doc_id` -> storage resolution.
- `handler/agent.go` + `agent_webhook.go`: detect `dataflow_canvas` and
run a sync debug returning chunks inline on the existing
chat/completions endpoint; reject DataFlow canvases from webhooks (fixes
the previously dead `== "DataFlow"` check; mirrors Python
`agent_api.py`).
- `parser_dispatch.go`: export `ParserFileFamily` for the executor's
page-cap injection.
### Debug run log + "View result"
Mirrors Python's debug-log contract so the front-end can replay each
component's progress and parsed output:
- `task/debug_log_sink.go`: a `DebugLogSink` records every component's
lifecycle into a `[{component_id, trace}]` array (each trace entry
carries `message`, `progress`, `timestamp`, `elapsed_time`). `Flush`
appends a terminal `END` marker whose first trace message is non-empty
so the front-end detects completion. On failure the END marker is
prefixed `[ERROR]` yet still carries the run, so the failure timeline
renders instead of being stuck empty. Timestamps and `elapsed_time` are
in seconds (matching the rest of the app).
- `task/debug_result_dsl.go`: `BuildDebugResultDSL` builds the `dsl` the
END marker carries — the Go analogue of Python's `Graph.__str__` +
END-marker `dsl` in `rag/flow/pipeline.py`. It combines the static DSL
structure (component_name / downstream / params / graph.nodes) with the
run output map (`output["state"][<id>]`) to emit, per component,
`obj.params.outputs[<format>].value` (chunks / text / json / html /
markdown) — the exact keys the front-end `dataflow-result` page reads to
render each step's parsed chunks. Raw embedding vectors (including the
dimension-scoped `q_<dim>_vec` keys) are stripped so the stored log
stays Python-scale.
- `task/pipeline_executor.go`: after the run, attach the built `dsl` to
the END marker via the `ResultSink` capability.
- `handler/agent.go`: `runCanvasPipelineDebug` generates a stable
`message_id` up-front and always flushes the log (success or failure);
`respondWithDebugResult` returns `message_id` in **both** the success
and the error envelope so the front-end can poll the log. The debug-log
endpoint `GET /agents/:id/logs/:message_id` serves the array.
- `web/src/pages/agent/hooks/use-run-dataflow.ts`: on a run failure,
also surface `message_id` via `setMessageId` so the log sheet renders
the failure timeline (the `[ERROR]` END marker is already written).
Guarded by `if (msgId)`, so it is a safe no-op when the back-end does
not return an id.
## Behavioral notes
- Debug parses only the first pages (`debugPageCapPages`) for a fast
preview; an explicit `pages` cap already present in the ParserConfig is
respected.
- Debug mode does not keep chunk images (see above) and does not compute
embeddings — it exercises parse + chunk only.
## Test plan
- Go: `debug_test.go`, `debug_log_sink_test.go` (trace pairing, END
marker, `[ERROR]` prefix, fractional-second timestamp/elapsed_time, size
caps, and a real-pipeline test asserting the END-marker `dsl` carries
non-empty per-component `params.outputs` with chunks),
`debug_result_dsl_test.go` (flat and real nested `output["state"]`
shapes, vector stripping, format priority),
`debug_pages_integration_test.go`, `pipeline_executor_persist_test.go`,
`handler/agent_pipeline_debug_test.go`, `handler/agent_logs_test.go`
(incl. `TestRunCanvasPipelineDebug_ErrorStillExposesMessageID` /
`TestRespondWithDebugResult_ErrorCarriesMessageID` locking `message_id`
on failure), plus updates to `agent_test.go` / `agent_webhook_test.go` /
`chunker/image_upload_test.go` / `tokenizer*.go`.
- `go build ./...` and `./build.sh --test` for affected packages.
🤖 Generated with [CodeBuddy Code](https://cnb.cool/codebuddy)
---------
Co-authored-by: CodeBuddy Code <noreply@tencent.com>
2026-07-31 13:10:08 +08:00
famEntry := parserConfig [ parserCpnID ] . ( map [ string ] any )
pdf := famEntry [ "pdf" ] . ( map [ string ] any )
if ! reflect . DeepEqual ( pdf [ "pages" ] , [ ] any { [ ] any { 1 , 1000000 } } ) {
t . Errorf ( "parserConfig[%q][\"pdf\"][\"pages\"] = %v, want [[1, 1000000]] (explicit cap must be respected, not overridden by debug default)" , parserCpnID , pdf [ "pages" ] )
}
} )
}
refactor(task): sink parser page-cap override into pipeline package (#17905)
## Summary
Moves the canvas-debug parser page-cap injection out of the `task`
orchestrator and into a **debug-agnostic** `pipeline` helper, so
`PipelineExecutor` keeps only the orchestration skeleton (resolving one
of the P1 review findings: the executor was overloaded with
DSL/parser-param assembly).
### Changes
- **`pipeline/parser_page_cap.go`** (new):
- `BuildParserPageCapOverride(parserConfig, dsl, docType, capPages int,
parserComponentName string, familyOf)` — injects the
`ParserConfig[cpnID][family]["pages"]` cap through the same
`override_params` channel production uses. The cap value and family
resolution are injected by the caller, so the function carries no debug
semantics and is reusable for any page-cap scenario.
- `ExtractParserCpnID(dsl, parserComponentName)` — shared Parser cpnID
discovery from (optionally enveloped) DSL.
- `UnwrapCanvasDSL(raw []byte)` — exported single source of truth for
stripping the `{"dsl": {...}}` canvas envelope.
- `pipeline` does **not** import `component` (no reverse dependency);
callers inject `component.ComponentNameParser` /
`component.ParserFileFamily`.
- **`task/pipeline_executor.go`**: removed `injectDebugPageCap` (the
`debugPageCapPages = 2` constant stays in the task package). The debug
branch now calls `pipeline.BuildParserPageCapOverride(...)`.
- **`task/pipeline_executor.go` `warnUnknownComponentParams`**: fixed a
production no-op bug — it passed the enveloped DSL straight to
`ExtractAllComponentParams`, which silently errored and disabled the
unknown-cpnID guard. It now unwraps the envelope first.
- **`task/debug_result_dsl.go`**: reuses `pipeline.UnwrapCanvasDSL`
instead of a third inline envelope-unwrap copy.
### Behavior
No external debug-preview behavior changes. The three original
invariants are preserved exactly:
1. explicit `pages` caps under `cpnID+family` are respected (not
overwritten),
2. an empty family (unknown docType) is a no-op,
3. the injected shape is `[]any{[]any{1, capPages}}` (the
`[]any`-of-`[]any` form `NormalizePDFPages` requires).
## Test plan
- New `pipeline/parser_page_cap_test.go`: `BuildParserPageCapOverride`
(inject / respect-existing / unknown-family no-op / no-Parser no-op),
`ExtractParserCpnID` (enveloped + raw), `UnwrapCanvasDSL`.
- `task/debug_test.go`: `TestInjectDebugPageCap` migrated to the new
helper; new
`TestWarnUnknownComponentParamsDetectsUnknownCPNFromEnvelope` captures
the warning via `zaptest/observer` to prove the envelope no-op bug is
fixed.
- Both `internal/ingestion/pipeline` and `internal/ingestion/task` pass
`build.sh --test` (unit tier).
## Notes
- `TOKEN_CHUNKER_HANDOFF.md` is an unrelated untracked file and was
deliberately **not** included in this PR.
2026-08-06 15:50:39 +08:00
// TestWarnUnknownComponentParamsDetectsUnknownCPNFromEnvelope pins the fix for
// the enveloped-DSL no-op bug: warnUnknownComponentParams previously passed the
// raw (enveloped) DSL straight to ExtractAllComponentParams, whose "components"
// key is nested under "dsl", so it errored and silently returned — never
// detecting unknown cpnIDs in production. The helper now unwraps the envelope
// first, so an unknown cpnID in parserConfig is actually surfaced.
func TestWarnUnknownComponentParamsDetectsUnknownCPNFromEnvelope ( t * testing . T ) {
core , recorded := observer . New ( zap . NewAtomicLevelAt ( zap . DebugLevel ) )
old := common . Logger
common . Logger = zap . New ( core )
defer func ( ) { common . Logger = old } ( )
// Enveloped DSL (production shape) carrying only a Parser component.
dsl := ` { "dsl": { "components": { "Parser:Abc": { "obj": { "component_name": "Parser", "params": { }}}}}} `
// parserConfig references a cpnID NOT present in the DSL -> must be warned.
parserConfig := map [ string ] any {
"Parser:Unknown" : map [ string ] any { "pdf" : map [ string ] any { } } ,
}
warnUnknownComponentParams ( dsl , parserConfig )
found := false
for _ , e := range recorded . All ( ) {
if strings . Contains ( e . Message , "Parser:Unknown" ) {
found = true
}
}
if ! found {
t . Fatalf ( "expected a warning about unknown cpnID Parser:Unknown (envelope DSL must be unwrapped); got logs: %v" , recorded . All ( ) )
}
}
// TestBuildDebugResultDSL_Envelope pins that BuildDebugResultDSL unwraps the
// canvas envelope before reading "components" — the same shared
// pipeline.UnwrapCanvasDSL the cap override and warnUnknownComponentParams
// use. An enveloped DSL (production shape {"dsl": {...}}) must resolve the
// components map exactly like the equivalent raw (non-enveloped) DSL.
func TestBuildDebugResultDSL_Envelope ( t * testing . T ) {
const compID = "Parser:Abc"
rawDSL := ` { "components": { " ` + compID + ` ": { "obj": { "component_name": "Parser", "params": { "parse_method": "deepdoc"}}}}} `
output := map [ string ] any {
"state" : map [ string ] any {
compID : map [ string ] any { "chunks" : [ ] any { map [ string ] any { "text" : "hi" } } } ,
} ,
}
// Raw (non-enveloped) DSL.
rawRes , err := BuildDebugResultDSL ( rawDSL , output )
if err != nil {
t . Fatalf ( "raw DSL: %v" , err )
}
rawComps , ok := rawRes [ "components" ] . ( map [ string ] any )
if ! ok {
t . Fatalf ( "raw DSL: components missing: %#v" , rawRes )
}
if _ , ok := rawComps [ compID ] ; ! ok {
t . Fatalf ( "raw DSL: components missing %q: %#v" , compID , rawComps )
}
// Enveloped DSL (production shape) must unwrap to the same result.
envDSL := ` { "dsl": ` + rawDSL + ` } `
envRes , err := BuildDebugResultDSL ( envDSL , output )
if err != nil {
t . Fatalf ( "enveloped DSL: %v" , err )
}
envComps , ok := envRes [ "components" ] . ( map [ string ] any )
if ! ok {
t . Fatalf ( "enveloped DSL: components missing (envelope not unwrapped?): %#v" , envRes )
}
if _ , ok := envComps [ compID ] ; ! ok {
t . Fatalf ( "enveloped DSL: components missing %q (envelope not unwrapped?): %#v" , compID , envComps )
}
// The two shapes must yield an identical component output.
if ! reflect . DeepEqual ( rawComps [ compID ] , envComps [ compID ] ) {
t . Fatalf ( "enveloped and raw DSL produced different results:\nraw=%#v\nenv=%#v" , rawComps [ compID ] , envComps [ compID ] )
}
}