Files

311 lines
11 KiB
Go
Raw Permalink Normal View History

feat(agent): Go port — canvas engine, 22 components, DSL v2, 13 endpoints (#15952) Ports the agent canvas subsystem from Python to Go. ## What's included ### Canvas Engine (Phase 0/1) - State engine, scheduler, variable resolver, Redis checkpoint store, cancel protocol - **209 tests** across canvas / component / io packages ### 22 Components (P0–P4) | Tier | Components | |---|---| | P0 T1+T2+T3 | LLM, Agent, ExitLoop, Switch, Categorize, Begin, Message, Invoke | | P1 T3 | VariableAggregator, VariableAssigner, StringTransform, ListOperations, DataOperations | | P2 T3 | Iteration, IterationItem, Loop, LoopItem | | P3 T3 | UserFillUp, Fillup | | P4 T5 | Browser, ExcelProcessor, DocsGenerator | ### DSL v2 Schema (Phase 2.5) - Typed v2 in-memory model with v1-to-v2 auto-detect converter - v1 legacy field stripping per plan §2.11.7 ### HTTP Endpoints & Bug Fixes (Plans PR1–PR3) - **DELETE SQL bug fix**: gorm v2 `Where("id = ?", id).Delete(...)` pattern - **CreateAgent validation**: title/DSL required, duplicate check, 103 envelope - **13 new endpoints**: templates, prompts, tags, sessions CRUD, chat/completions (SSE + non-stream stubs), rerun, test_db_connection, logs, webhook/logs - **756 Go unit tests** (745 → 756, +18) - **17 → 0 Python integration test failures** (test_agents.py + test_session_management/) ### Tools 21 eino tools: HTTPHelper, search tools, financial/data tools, mandatory stubs ### Infrastructure OTel observability, NATS message queue, DeepDoc gRPC client, SSRF guards, IDOR mitigation
2026-06-12 22:58:28 +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 tool
import (
"context"
"encoding/json"
"errors"
"fmt"
"os"
feat(agent): Go port — canvas engine, 22 components, DSL v2, 13 endpoints (#15952) Ports the agent canvas subsystem from Python to Go. ## What's included ### Canvas Engine (Phase 0/1) - State engine, scheduler, variable resolver, Redis checkpoint store, cancel protocol - **209 tests** across canvas / component / io packages ### 22 Components (P0–P4) | Tier | Components | |---|---| | P0 T1+T2+T3 | LLM, Agent, ExitLoop, Switch, Categorize, Begin, Message, Invoke | | P1 T3 | VariableAggregator, VariableAssigner, StringTransform, ListOperations, DataOperations | | P2 T3 | Iteration, IterationItem, Loop, LoopItem | | P3 T3 | UserFillUp, Fillup | | P4 T5 | Browser, ExcelProcessor, DocsGenerator | ### DSL v2 Schema (Phase 2.5) - Typed v2 in-memory model with v1-to-v2 auto-detect converter - v1 legacy field stripping per plan §2.11.7 ### HTTP Endpoints & Bug Fixes (Plans PR1–PR3) - **DELETE SQL bug fix**: gorm v2 `Where("id = ?", id).Delete(...)` pattern - **CreateAgent validation**: title/DSL required, duplicate check, 103 envelope - **13 new endpoints**: templates, prompts, tags, sessions CRUD, chat/completions (SSE + non-stream stubs), rerun, test_db_connection, logs, webhook/logs - **756 Go unit tests** (745 → 756, +18) - **17 → 0 Python integration test failures** (test_agents.py + test_session_management/) ### Tools 21 eino tools: HTTPHelper, search tools, financial/data tools, mandatory stubs ### Infrastructure OTel observability, NATS message queue, DeepDoc gRPC client, SSRF guards, IDOR mitigation
2026-06-12 22:58:28 +08:00
"strings"
"github.com/cloudwego/eino/components/tool"
"github.com/cloudwego/eino/schema"
"go.uber.org/zap"
"ragflow/internal/common"
feat(agent): Go port — canvas engine, 22 components, DSL v2, 13 endpoints (#15952) Ports the agent canvas subsystem from Python to Go. ## What's included ### Canvas Engine (Phase 0/1) - State engine, scheduler, variable resolver, Redis checkpoint store, cancel protocol - **209 tests** across canvas / component / io packages ### 22 Components (P0–P4) | Tier | Components | |---|---| | P0 T1+T2+T3 | LLM, Agent, ExitLoop, Switch, Categorize, Begin, Message, Invoke | | P1 T3 | VariableAggregator, VariableAssigner, StringTransform, ListOperations, DataOperations | | P2 T3 | Iteration, IterationItem, Loop, LoopItem | | P3 T3 | UserFillUp, Fillup | | P4 T5 | Browser, ExcelProcessor, DocsGenerator | ### DSL v2 Schema (Phase 2.5) - Typed v2 in-memory model with v1-to-v2 auto-detect converter - v1 legacy field stripping per plan §2.11.7 ### HTTP Endpoints & Bug Fixes (Plans PR1–PR3) - **DELETE SQL bug fix**: gorm v2 `Where("id = ?", id).Delete(...)` pattern - **CreateAgent validation**: title/DSL required, duplicate check, 103 envelope - **13 new endpoints**: templates, prompts, tags, sessions CRUD, chat/completions (SSE + non-stream stubs), rerun, test_db_connection, logs, webhook/logs - **756 Go unit tests** (745 → 756, +18) - **17 → 0 Python integration test failures** (test_agents.py + test_session_management/) ### Tools 21 eino tools: HTTPHelper, search tools, financial/data tools, mandatory stubs ### Infrastructure OTel observability, NATS message queue, DeepDoc gRPC client, SSRF guards, IDOR mitigation
2026-06-12 22:58:28 +08:00
)
// ErrCodeExecSandboxMissing is returned when no sandbox client is
// registered. The Python sandbox itself is kept as-is (the Go
// side never reimplemented the sandbox). When a client is
// registered via SetSandboxClient at boot, the tool dispatches
// the execution.
feat(agent): Go port — canvas engine, 22 components, DSL v2, 13 endpoints (#15952) Ports the agent canvas subsystem from Python to Go. ## What's included ### Canvas Engine (Phase 0/1) - State engine, scheduler, variable resolver, Redis checkpoint store, cancel protocol - **209 tests** across canvas / component / io packages ### 22 Components (P0–P4) | Tier | Components | |---|---| | P0 T1+T2+T3 | LLM, Agent, ExitLoop, Switch, Categorize, Begin, Message, Invoke | | P1 T3 | VariableAggregator, VariableAssigner, StringTransform, ListOperations, DataOperations | | P2 T3 | Iteration, IterationItem, Loop, LoopItem | | P3 T3 | UserFillUp, Fillup | | P4 T5 | Browser, ExcelProcessor, DocsGenerator | ### DSL v2 Schema (Phase 2.5) - Typed v2 in-memory model with v1-to-v2 auto-detect converter - v1 legacy field stripping per plan §2.11.7 ### HTTP Endpoints & Bug Fixes (Plans PR1–PR3) - **DELETE SQL bug fix**: gorm v2 `Where("id = ?", id).Delete(...)` pattern - **CreateAgent validation**: title/DSL required, duplicate check, 103 envelope - **13 new endpoints**: templates, prompts, tags, sessions CRUD, chat/completions (SSE + non-stream stubs), rerun, test_db_connection, logs, webhook/logs - **756 Go unit tests** (745 → 756, +18) - **17 → 0 Python integration test failures** (test_agents.py + test_session_management/) ### Tools 21 eino tools: HTTPHelper, search tools, financial/data tools, mandatory stubs ### Infrastructure OTel observability, NATS message queue, DeepDoc gRPC client, SSRF guards, IDOR mitigation
2026-06-12 22:58:28 +08:00
var ErrCodeExecSandboxMissing = errors.New(
"CodeExec sandbox client not registered — call SetSandboxClient at boot",
feat(agent): Go port — canvas engine, 22 components, DSL v2, 13 endpoints (#15952) Ports the agent canvas subsystem from Python to Go. ## What's included ### Canvas Engine (Phase 0/1) - State engine, scheduler, variable resolver, Redis checkpoint store, cancel protocol - **209 tests** across canvas / component / io packages ### 22 Components (P0–P4) | Tier | Components | |---|---| | P0 T1+T2+T3 | LLM, Agent, ExitLoop, Switch, Categorize, Begin, Message, Invoke | | P1 T3 | VariableAggregator, VariableAssigner, StringTransform, ListOperations, DataOperations | | P2 T3 | Iteration, IterationItem, Loop, LoopItem | | P3 T3 | UserFillUp, Fillup | | P4 T5 | Browser, ExcelProcessor, DocsGenerator | ### DSL v2 Schema (Phase 2.5) - Typed v2 in-memory model with v1-to-v2 auto-detect converter - v1 legacy field stripping per plan §2.11.7 ### HTTP Endpoints & Bug Fixes (Plans PR1–PR3) - **DELETE SQL bug fix**: gorm v2 `Where("id = ?", id).Delete(...)` pattern - **CreateAgent validation**: title/DSL required, duplicate check, 103 envelope - **13 new endpoints**: templates, prompts, tags, sessions CRUD, chat/completions (SSE + non-stream stubs), rerun, test_db_connection, logs, webhook/logs - **756 Go unit tests** (745 → 756, +18) - **17 → 0 Python integration test failures** (test_agents.py + test_session_management/) ### Tools 21 eino tools: HTTPHelper, search tools, financial/data tools, mandatory stubs ### Infrastructure OTel observability, NATS message queue, DeepDoc gRPC client, SSRF guards, IDOR mitigation
2026-06-12 22:58:28 +08:00
)
const codeExecToolName = "execute_code"
const codeExecToolDescription = "This tool has a sandbox that can execute code written in 'Python'/'Javascript'. " +
"It receives a piece of code and returns a JSON string."
// codeExecArgs is the JSON shape the model sends in. The Python
// tool accepts "lang" + "script"; we also accept "code" as a
// synonym since some DSLs and tests use that spelling.
feat(agent): Go port — canvas engine, 22 components, DSL v2, 13 endpoints (#15952) Ports the agent canvas subsystem from Python to Go. ## What's included ### Canvas Engine (Phase 0/1) - State engine, scheduler, variable resolver, Redis checkpoint store, cancel protocol - **209 tests** across canvas / component / io packages ### 22 Components (P0–P4) | Tier | Components | |---|---| | P0 T1+T2+T3 | LLM, Agent, ExitLoop, Switch, Categorize, Begin, Message, Invoke | | P1 T3 | VariableAggregator, VariableAssigner, StringTransform, ListOperations, DataOperations | | P2 T3 | Iteration, IterationItem, Loop, LoopItem | | P3 T3 | UserFillUp, Fillup | | P4 T5 | Browser, ExcelProcessor, DocsGenerator | ### DSL v2 Schema (Phase 2.5) - Typed v2 in-memory model with v1-to-v2 auto-detect converter - v1 legacy field stripping per plan §2.11.7 ### HTTP Endpoints & Bug Fixes (Plans PR1–PR3) - **DELETE SQL bug fix**: gorm v2 `Where("id = ?", id).Delete(...)` pattern - **CreateAgent validation**: title/DSL required, duplicate check, 103 envelope - **13 new endpoints**: templates, prompts, tags, sessions CRUD, chat/completions (SSE + non-stream stubs), rerun, test_db_connection, logs, webhook/logs - **756 Go unit tests** (745 → 756, +18) - **17 → 0 Python integration test failures** (test_agents.py + test_session_management/) ### Tools 21 eino tools: HTTPHelper, search tools, financial/data tools, mandatory stubs ### Infrastructure OTel observability, NATS message queue, DeepDoc gRPC client, SSRF guards, IDOR mitigation
2026-06-12 22:58:28 +08:00
type codeExecArgs struct {
Language string `json:"language,omitempty"`
Lang string `json:"lang,omitempty"`
Script string `json:"script,omitempty"`
Code string `json:"code,omitempty"`
Args map[string]any `json:"arguments,omitempty"`
// Timeout is the per-execution wall-clock budget in seconds. 0
// (the default) defers to the sandbox provider's own default
// (typically 30s). Mirrors Python's
// `code_exec.py:358 timeout_seconds = int(os.environ.get(...))`
// but the value flows per-call rather than per-process, which
// lets the model dial up/down for known-fast vs. known-slow
// scripts.
Timeout int `json:"timeout,omitempty"`
feat(agent): Go port — canvas engine, 22 components, DSL v2, 13 endpoints (#15952) Ports the agent canvas subsystem from Python to Go. ## What's included ### Canvas Engine (Phase 0/1) - State engine, scheduler, variable resolver, Redis checkpoint store, cancel protocol - **209 tests** across canvas / component / io packages ### 22 Components (P0–P4) | Tier | Components | |---|---| | P0 T1+T2+T3 | LLM, Agent, ExitLoop, Switch, Categorize, Begin, Message, Invoke | | P1 T3 | VariableAggregator, VariableAssigner, StringTransform, ListOperations, DataOperations | | P2 T3 | Iteration, IterationItem, Loop, LoopItem | | P3 T3 | UserFillUp, Fillup | | P4 T5 | Browser, ExcelProcessor, DocsGenerator | ### DSL v2 Schema (Phase 2.5) - Typed v2 in-memory model with v1-to-v2 auto-detect converter - v1 legacy field stripping per plan §2.11.7 ### HTTP Endpoints & Bug Fixes (Plans PR1–PR3) - **DELETE SQL bug fix**: gorm v2 `Where("id = ?", id).Delete(...)` pattern - **CreateAgent validation**: title/DSL required, duplicate check, 103 envelope - **13 new endpoints**: templates, prompts, tags, sessions CRUD, chat/completions (SSE + non-stream stubs), rerun, test_db_connection, logs, webhook/logs - **756 Go unit tests** (745 → 756, +18) - **17 → 0 Python integration test failures** (test_agents.py + test_session_management/) ### Tools 21 eino tools: HTTPHelper, search tools, financial/data tools, mandatory stubs ### Infrastructure OTel observability, NATS message queue, DeepDoc gRPC client, SSRF guards, IDOR mitigation
2026-06-12 22:58:28 +08:00
}
// codeExecResult is the JSON envelope returned to the model. The output
// shape mirrors the Python tool's `content` / `_ERROR` / `actual_type`
// fields so downstream nodes can pattern-match unchanged. Artifacts and
// Attachments are surfaced for the model and downstream component
// consumption (e.g. Message component's artifact markdown formatter).
feat(agent): Go port — canvas engine, 22 components, DSL v2, 13 endpoints (#15952) Ports the agent canvas subsystem from Python to Go. ## What's included ### Canvas Engine (Phase 0/1) - State engine, scheduler, variable resolver, Redis checkpoint store, cancel protocol - **209 tests** across canvas / component / io packages ### 22 Components (P0–P4) | Tier | Components | |---|---| | P0 T1+T2+T3 | LLM, Agent, ExitLoop, Switch, Categorize, Begin, Message, Invoke | | P1 T3 | VariableAggregator, VariableAssigner, StringTransform, ListOperations, DataOperations | | P2 T3 | Iteration, IterationItem, Loop, LoopItem | | P3 T3 | UserFillUp, Fillup | | P4 T5 | Browser, ExcelProcessor, DocsGenerator | ### DSL v2 Schema (Phase 2.5) - Typed v2 in-memory model with v1-to-v2 auto-detect converter - v1 legacy field stripping per plan §2.11.7 ### HTTP Endpoints & Bug Fixes (Plans PR1–PR3) - **DELETE SQL bug fix**: gorm v2 `Where("id = ?", id).Delete(...)` pattern - **CreateAgent validation**: title/DSL required, duplicate check, 103 envelope - **13 new endpoints**: templates, prompts, tags, sessions CRUD, chat/completions (SSE + non-stream stubs), rerun, test_db_connection, logs, webhook/logs - **756 Go unit tests** (745 → 756, +18) - **17 → 0 Python integration test failures** (test_agents.py + test_session_management/) ### Tools 21 eino tools: HTTPHelper, search tools, financial/data tools, mandatory stubs ### Infrastructure OTel observability, NATS message queue, DeepDoc gRPC client, SSRF guards, IDOR mitigation
2026-06-12 22:58:28 +08:00
type codeExecResult struct {
Content string `json:"content,omitempty"`
ActualType string `json:"actual_type,omitempty"`
feat(agent): align Go agent behavior with Python (except retrieval component) (#16225) ## Summary Aligns the **Go agent runtime/canvas/components/tools** behavior with the **Python `agent/` implementation** so the same stored canvas DSL produces the same execution result on either side. Every component, tool, and runtime primitive in `internal/agent/` is now driven by the same semantics as its Python counterpart — variable resolution, template substitution, control flow, error reporting, retry/cancel, and stream event shapes. The **retrieval component is the one explicit exception** in this PR. It is being reworked in a separate change and is excluded from this alignment pass; the wrapper slot (`universe_a_wrappers.go → newRetrievalComponent`) is preserved. ## Scope of alignment ### Components (all aligned with `agent/component/`) `Begin` · `Message` · `LLM` (incl. ChatTemplateKwargs, MessageHistoryWindowSize, VisualFiles, Cite, OutputStructure, JSONOutput, TopP, MaxRetries, DelayAfterError, credentials) · `Agent` (react + tool artifact capture + `Reset()` interface-assert) · `Switch` (12/12 operators, Python-equivalent semantics) · `Categorize` · `Invoke` · `Iteration` · `Loop` (macro-expansion through `workflowx.AddLoopNode`) · `UserFillUp` (Python-equivalent interrupt/resume via eino `compose.Interrupt`/`ResumeWithData`) · `FillUp` · `DataOperations` · `ListOperations` · `StringTransform` · `VariableAggregator` · `VariableAssigner` · `Browser` (full stagehand runtime parity) · `DocsGenerator` · `ExcelProcessor`. ### Tools (all aligned with `agent/tools/`) `Retrieval` (wrapper slot only — logic out of scope) · `MCPToolAdapter` (streamable-HTTP) · `CodeExec` (sandbox bridge with `code_exec_contract.go` matching Python contract) · `AkShare` · `ArXiv` · `Crawler` · `DeepL` · `DuckDuckGo` · `Email` · `ExeSQL` · `GitHub` · `Google` · `GoogleScholar` · `Jin10` · `PubMed` · `QWeather` · `SearXNG` · `Tavily` · `Tushare` · `Wencai` · `Wikipedia` · `YahooFinance` — uniform `eino tool.InvokableTool` interface, SSRF protection, shared HTTP client. ### Canvas execution engine (`internal/agent/canvas/`) Aligned with Python's `agent/canvas.py`: - **Scheduler** (`scheduler.go`): state pre/post handlers, node lambdas, per-component timeout resolver (4-level: per-class env → per-class table → uniform env → 600s fallback), `legacyNoOpNames`. - **Loop subgraph** (`loop_subgraph.go`): Python-equivalent `AddLoopNode` macro expansion + condition translation. - **Multibranch** (`multibranch.go`): `Switch` / `Categorize` routing via `compose.NewGraphMultiBranch` — same branch selection semantics as Python. - **Parallel subgraph** (`parallel_subgraph.go`): matches Python's parallel fan-out contract. - **Interrupt/Resume** (`interrupt_resume.go`): `UserFillUpNodeBody` / `IsInterruptError` / `ExtractInterruptContexts` — replaces the deprecated Python sentinel chain with eino's native interrupt API, preserving the same external behavior. - **Checkpoint** (`checkpoint_store.go`): `RedisCheckPointStore` Get/Set/Delete, with business metadata (status / canvas_id / parent_run_id) on a parallel Redis Hash. - **RunTracker** (`run_tracker.go`): Start / MarkSucceeded / MarkFailed / MarkCancelled / AttachCheckpoint — same lifecycle as the Python run record. - **Cancel** (`cancel.go`): Redis pub/sub watch. - **Stream** (`stream.go`): SSE channel with `messages` / `waiting` / `errors` / `done` events, same shape as Python's `agent.canvas.RunEvent` payload. ### DSL bridge (`internal/agent/dsl/`) - `normalize.go`: v1↔v2 collapsed into a single wire format — Python and Go consume the same stored JSON. - `reset.go`: per-run state reset matches Python's `Canvas.reset()` semantics. - Testdata mirrors Python's `agent_msg.json` / `all.json` / etc. ### Runtime (`internal/agent/runtime/`) - `CanvasState` / `NewCanvasState` / `GetVar` / `SetVar` / `ReadVars`: same `{{cpn_id@param}}` resolution model. - `ResolveTemplate` (regex fast path + gonja fallback) — Python Jinja-style semantics. - `selector.go`, `metrics.go`, `component.go`: shared runtime contracts. ## Out of scope (intentionally) - **`Retrieval` component logic** — wrapped only; full parity lands in a follow-up PR. - **Frontend** — only minor dsl-bridge / canvas UX fixes ride along. - **CLI / admin / model registry** — orthogonal to agent behavior. ## How alignment is verified `internal/service/agent_run_e2e_test.go` exercises the **full production chain** against real Python-shaped DSL fixtures: ``` loadCanvasForUser → versionDAO.GetLatest → decodeCanvasFromDSL → canvas.Compile → cc.Workflow.Invoke → answer extraction ``` using in-memory SQLite + miniredis (no Docker). Covers: - `TestRunAgent_RealCanvas_BeginMessage` — happy path, `{{sys.query}}` resolution - `TestRunAgent_RealCanvas_WaitForUserResume` — two-run resume cycle (Python-equivalent) - `TestRunAgent_RealCanvas_CompileFails` — unknown component name → sanitized error (Python-equivalent) - `TestRunAgent_RealCanvas_InvokeFails` — unresolvable template ref (Python-equivalent) - `TestRunAgent_RunTracker_AttachCheckpoint_CallSequence` — Start→AttachCheckpoint→MarkSucceeded lifecycle `internal/handler/agent_test.go` — SSE streaming parity (`Content-Type: text/event-stream`, `data: {…}\n\n`, trailing `data: [DONE]\n\n`, OpenAI-compatible non-stream `choices`). `internal/agent/canvas/fixture_compile_test.go` + per-component tests pin the Python-equivalent outputs. ``` go test -count=1 -v -run 'TestRunAgent_RealCanvas|TestRunAgent_RunTracker' ./internal/service/ ``` ## Design reference `docs/develop/agent-go-port-design.md` (1329 lines, last cross-checked 2026-06-17) — module layout, per-component / per-tool inventory, corner-case catalogue, and the actionable backlog (Section 14, including the retrieval alignment follow-up). --------- Co-authored-by: Claude <noreply@anthropic.com>
2026-06-22 11:58:29 +08:00
RawResult any `json:"raw_result,omitempty"`
Stub bool `json:"stub,omitempty"`
Error string `json:"_ERROR,omitempty"`
ExitCode int `json:"exit_code,omitempty"`
Stdout string `json:"stdout,omitempty"`
Stderr string `json:"stderr,omitempty"`
Artifacts []map[string]any `json:"_ARTIFACTS,omitempty"`
Attachments []map[string]any `json:"attachments,omitempty"`
feat(agent): Go port — canvas engine, 22 components, DSL v2, 13 endpoints (#15952) Ports the agent canvas subsystem from Python to Go. ## What's included ### Canvas Engine (Phase 0/1) - State engine, scheduler, variable resolver, Redis checkpoint store, cancel protocol - **209 tests** across canvas / component / io packages ### 22 Components (P0–P4) | Tier | Components | |---|---| | P0 T1+T2+T3 | LLM, Agent, ExitLoop, Switch, Categorize, Begin, Message, Invoke | | P1 T3 | VariableAggregator, VariableAssigner, StringTransform, ListOperations, DataOperations | | P2 T3 | Iteration, IterationItem, Loop, LoopItem | | P3 T3 | UserFillUp, Fillup | | P4 T5 | Browser, ExcelProcessor, DocsGenerator | ### DSL v2 Schema (Phase 2.5) - Typed v2 in-memory model with v1-to-v2 auto-detect converter - v1 legacy field stripping per plan §2.11.7 ### HTTP Endpoints & Bug Fixes (Plans PR1–PR3) - **DELETE SQL bug fix**: gorm v2 `Where("id = ?", id).Delete(...)` pattern - **CreateAgent validation**: title/DSL required, duplicate check, 103 envelope - **13 new endpoints**: templates, prompts, tags, sessions CRUD, chat/completions (SSE + non-stream stubs), rerun, test_db_connection, logs, webhook/logs - **756 Go unit tests** (745 → 756, +18) - **17 → 0 Python integration test failures** (test_agents.py + test_session_management/) ### Tools 21 eino tools: HTTPHelper, search tools, financial/data tools, mandatory stubs ### Infrastructure OTel observability, NATS message queue, DeepDoc gRPC client, SSRF guards, IDOR mitigation
2026-06-12 22:58:28 +08:00
}
// CodeExecTool is the for the CodeExec tool
// ( . It validates language +
feat(agent): Go port — canvas engine, 22 components, DSL v2, 13 endpoints (#15952) Ports the agent canvas subsystem from Python to Go. ## What's included ### Canvas Engine (Phase 0/1) - State engine, scheduler, variable resolver, Redis checkpoint store, cancel protocol - **209 tests** across canvas / component / io packages ### 22 Components (P0–P4) | Tier | Components | |---|---| | P0 T1+T2+T3 | LLM, Agent, ExitLoop, Switch, Categorize, Begin, Message, Invoke | | P1 T3 | VariableAggregator, VariableAssigner, StringTransform, ListOperations, DataOperations | | P2 T3 | Iteration, IterationItem, Loop, LoopItem | | P3 T3 | UserFillUp, Fillup | | P4 T5 | Browser, ExcelProcessor, DocsGenerator | ### DSL v2 Schema (Phase 2.5) - Typed v2 in-memory model with v1-to-v2 auto-detect converter - v1 legacy field stripping per plan §2.11.7 ### HTTP Endpoints & Bug Fixes (Plans PR1–PR3) - **DELETE SQL bug fix**: gorm v2 `Where("id = ?", id).Delete(...)` pattern - **CreateAgent validation**: title/DSL required, duplicate check, 103 envelope - **13 new endpoints**: templates, prompts, tags, sessions CRUD, chat/completions (SSE + non-stream stubs), rerun, test_db_connection, logs, webhook/logs - **756 Go unit tests** (745 → 756, +18) - **17 → 0 Python integration test failures** (test_agents.py + test_session_management/) ### Tools 21 eino tools: HTTPHelper, search tools, financial/data tools, mandatory stubs ### Infrastructure OTel observability, NATS message queue, DeepDoc gRPC client, SSRF guards, IDOR mitigation
2026-06-12 22:58:28 +08:00
// non-empty code and returns a structured "not-yet-wired" error.
type CodeExecTool struct{}
// NewCodeExecTool returns a CodeExecTool implementing eino's
// tool.InvokableTool interface.
func NewCodeExecTool() *CodeExecTool {
return &CodeExecTool{}
}
// Info returns the tool's metadata for the chat model. The schema mirrors
// the Python CodeExecParam ToolMeta (plan , 字段对齐).
feat(agent): Go port — canvas engine, 22 components, DSL v2, 13 endpoints (#15952) Ports the agent canvas subsystem from Python to Go. ## What's included ### Canvas Engine (Phase 0/1) - State engine, scheduler, variable resolver, Redis checkpoint store, cancel protocol - **209 tests** across canvas / component / io packages ### 22 Components (P0–P4) | Tier | Components | |---|---| | P0 T1+T2+T3 | LLM, Agent, ExitLoop, Switch, Categorize, Begin, Message, Invoke | | P1 T3 | VariableAggregator, VariableAssigner, StringTransform, ListOperations, DataOperations | | P2 T3 | Iteration, IterationItem, Loop, LoopItem | | P3 T3 | UserFillUp, Fillup | | P4 T5 | Browser, ExcelProcessor, DocsGenerator | ### DSL v2 Schema (Phase 2.5) - Typed v2 in-memory model with v1-to-v2 auto-detect converter - v1 legacy field stripping per plan §2.11.7 ### HTTP Endpoints & Bug Fixes (Plans PR1–PR3) - **DELETE SQL bug fix**: gorm v2 `Where("id = ?", id).Delete(...)` pattern - **CreateAgent validation**: title/DSL required, duplicate check, 103 envelope - **13 new endpoints**: templates, prompts, tags, sessions CRUD, chat/completions (SSE + non-stream stubs), rerun, test_db_connection, logs, webhook/logs - **756 Go unit tests** (745 → 756, +18) - **17 → 0 Python integration test failures** (test_agents.py + test_session_management/) ### Tools 21 eino tools: HTTPHelper, search tools, financial/data tools, mandatory stubs ### Infrastructure OTel observability, NATS message queue, DeepDoc gRPC client, SSRF guards, IDOR mitigation
2026-06-12 22:58:28 +08:00
func (c *CodeExecTool) Info(_ context.Context) (*schema.ToolInfo, error) {
return &schema.ToolInfo{
Name: codeExecToolName,
Desc: codeExecToolDescription,
ParamsOneOf: schema.NewParamsOneOfByParams(map[string]*schema.ParameterInfo{
"language": {
Type: schema.String,
Desc: "The programming language of the code. Allowed: 'python' (or 'python3'), 'javascript' (or 'nodejs').",
Enum: []string{"python", "python3", "javascript", "nodejs"},
Required: true,
},
"code": {
Type: schema.String,
Desc: "The code to execute. Must define a `main` function (Python) or export `main` (JavaScript).",
Required: true,
},
}),
}, nil
}
// InvokableRun validates the inputs and dispatches to the
// registered sandbox client via SetSandboxClient.
feat(agent): Go port — canvas engine, 22 components, DSL v2, 13 endpoints (#15952) Ports the agent canvas subsystem from Python to Go. ## What's included ### Canvas Engine (Phase 0/1) - State engine, scheduler, variable resolver, Redis checkpoint store, cancel protocol - **209 tests** across canvas / component / io packages ### 22 Components (P0–P4) | Tier | Components | |---|---| | P0 T1+T2+T3 | LLM, Agent, ExitLoop, Switch, Categorize, Begin, Message, Invoke | | P1 T3 | VariableAggregator, VariableAssigner, StringTransform, ListOperations, DataOperations | | P2 T3 | Iteration, IterationItem, Loop, LoopItem | | P3 T3 | UserFillUp, Fillup | | P4 T5 | Browser, ExcelProcessor, DocsGenerator | ### DSL v2 Schema (Phase 2.5) - Typed v2 in-memory model with v1-to-v2 auto-detect converter - v1 legacy field stripping per plan §2.11.7 ### HTTP Endpoints & Bug Fixes (Plans PR1–PR3) - **DELETE SQL bug fix**: gorm v2 `Where("id = ?", id).Delete(...)` pattern - **CreateAgent validation**: title/DSL required, duplicate check, 103 envelope - **13 new endpoints**: templates, prompts, tags, sessions CRUD, chat/completions (SSE + non-stream stubs), rerun, test_db_connection, logs, webhook/logs - **756 Go unit tests** (745 → 756, +18) - **17 → 0 Python integration test failures** (test_agents.py + test_session_management/) ### Tools 21 eino tools: HTTPHelper, search tools, financial/data tools, mandatory stubs ### Infrastructure OTel observability, NATS message queue, DeepDoc gRPC client, SSRF guards, IDOR mitigation
2026-06-12 22:58:28 +08:00
func (c *CodeExecTool) InvokableRun(ctx context.Context, argumentsInJSON string, _ ...tool.Option) (string, error) {
var args codeExecArgs
if argumentsInJSON == "" {
return codeExecStubResult("arguments are required"), errors.New("code_exec: empty arguments")
}
if err := json.Unmarshal([]byte(argumentsInJSON), &args); err != nil {
return codeExecStubResult("invalid JSON: " + err.Error()),
fmt.Errorf("code_exec: parse arguments: %w", err)
}
lang := normalizeCodeExecLang(args.Language, args.Lang)
if lang == "" {
return codeExecStubResult("unsupported language: must be 'python'/'python3'/'javascript'/'nodejs'"),
errors.New("code_exec: invalid language")
}
script := args.Script
if script == "" {
script = args.Code
}
if strings.TrimSpace(script) == "" {
return codeExecStubResult("code is required"), errors.New("code_exec: empty code")
}
// Dispatch to the registered SandboxClient. When the default
// stub is in place, the call surfaces
// ErrCodeExecSandboxMissing; once a real client is
// installed via SetSandboxClient at boot, the script runs.
client := GetSandboxClient()
feat(agent): align Go agent behavior with Python (except retrieval component) (#16225) ## Summary Aligns the **Go agent runtime/canvas/components/tools** behavior with the **Python `agent/` implementation** so the same stored canvas DSL produces the same execution result on either side. Every component, tool, and runtime primitive in `internal/agent/` is now driven by the same semantics as its Python counterpart — variable resolution, template substitution, control flow, error reporting, retry/cancel, and stream event shapes. The **retrieval component is the one explicit exception** in this PR. It is being reworked in a separate change and is excluded from this alignment pass; the wrapper slot (`universe_a_wrappers.go → newRetrievalComponent`) is preserved. ## Scope of alignment ### Components (all aligned with `agent/component/`) `Begin` · `Message` · `LLM` (incl. ChatTemplateKwargs, MessageHistoryWindowSize, VisualFiles, Cite, OutputStructure, JSONOutput, TopP, MaxRetries, DelayAfterError, credentials) · `Agent` (react + tool artifact capture + `Reset()` interface-assert) · `Switch` (12/12 operators, Python-equivalent semantics) · `Categorize` · `Invoke` · `Iteration` · `Loop` (macro-expansion through `workflowx.AddLoopNode`) · `UserFillUp` (Python-equivalent interrupt/resume via eino `compose.Interrupt`/`ResumeWithData`) · `FillUp` · `DataOperations` · `ListOperations` · `StringTransform` · `VariableAggregator` · `VariableAssigner` · `Browser` (full stagehand runtime parity) · `DocsGenerator` · `ExcelProcessor`. ### Tools (all aligned with `agent/tools/`) `Retrieval` (wrapper slot only — logic out of scope) · `MCPToolAdapter` (streamable-HTTP) · `CodeExec` (sandbox bridge with `code_exec_contract.go` matching Python contract) · `AkShare` · `ArXiv` · `Crawler` · `DeepL` · `DuckDuckGo` · `Email` · `ExeSQL` · `GitHub` · `Google` · `GoogleScholar` · `Jin10` · `PubMed` · `QWeather` · `SearXNG` · `Tavily` · `Tushare` · `Wencai` · `Wikipedia` · `YahooFinance` — uniform `eino tool.InvokableTool` interface, SSRF protection, shared HTTP client. ### Canvas execution engine (`internal/agent/canvas/`) Aligned with Python's `agent/canvas.py`: - **Scheduler** (`scheduler.go`): state pre/post handlers, node lambdas, per-component timeout resolver (4-level: per-class env → per-class table → uniform env → 600s fallback), `legacyNoOpNames`. - **Loop subgraph** (`loop_subgraph.go`): Python-equivalent `AddLoopNode` macro expansion + condition translation. - **Multibranch** (`multibranch.go`): `Switch` / `Categorize` routing via `compose.NewGraphMultiBranch` — same branch selection semantics as Python. - **Parallel subgraph** (`parallel_subgraph.go`): matches Python's parallel fan-out contract. - **Interrupt/Resume** (`interrupt_resume.go`): `UserFillUpNodeBody` / `IsInterruptError` / `ExtractInterruptContexts` — replaces the deprecated Python sentinel chain with eino's native interrupt API, preserving the same external behavior. - **Checkpoint** (`checkpoint_store.go`): `RedisCheckPointStore` Get/Set/Delete, with business metadata (status / canvas_id / parent_run_id) on a parallel Redis Hash. - **RunTracker** (`run_tracker.go`): Start / MarkSucceeded / MarkFailed / MarkCancelled / AttachCheckpoint — same lifecycle as the Python run record. - **Cancel** (`cancel.go`): Redis pub/sub watch. - **Stream** (`stream.go`): SSE channel with `messages` / `waiting` / `errors` / `done` events, same shape as Python's `agent.canvas.RunEvent` payload. ### DSL bridge (`internal/agent/dsl/`) - `normalize.go`: v1↔v2 collapsed into a single wire format — Python and Go consume the same stored JSON. - `reset.go`: per-run state reset matches Python's `Canvas.reset()` semantics. - Testdata mirrors Python's `agent_msg.json` / `all.json` / etc. ### Runtime (`internal/agent/runtime/`) - `CanvasState` / `NewCanvasState` / `GetVar` / `SetVar` / `ReadVars`: same `{{cpn_id@param}}` resolution model. - `ResolveTemplate` (regex fast path + gonja fallback) — Python Jinja-style semantics. - `selector.go`, `metrics.go`, `component.go`: shared runtime contracts. ## Out of scope (intentionally) - **`Retrieval` component logic** — wrapped only; full parity lands in a follow-up PR. - **Frontend** — only minor dsl-bridge / canvas UX fixes ride along. - **CLI / admin / model registry** — orthogonal to agent behavior. ## How alignment is verified `internal/service/agent_run_e2e_test.go` exercises the **full production chain** against real Python-shaped DSL fixtures: ``` loadCanvasForUser → versionDAO.GetLatest → decodeCanvasFromDSL → canvas.Compile → cc.Workflow.Invoke → answer extraction ``` using in-memory SQLite + miniredis (no Docker). Covers: - `TestRunAgent_RealCanvas_BeginMessage` — happy path, `{{sys.query}}` resolution - `TestRunAgent_RealCanvas_WaitForUserResume` — two-run resume cycle (Python-equivalent) - `TestRunAgent_RealCanvas_CompileFails` — unknown component name → sanitized error (Python-equivalent) - `TestRunAgent_RealCanvas_InvokeFails` — unresolvable template ref (Python-equivalent) - `TestRunAgent_RunTracker_AttachCheckpoint_CallSequence` — Start→AttachCheckpoint→MarkSucceeded lifecycle `internal/handler/agent_test.go` — SSE streaming parity (`Content-Type: text/event-stream`, `data: {…}\n\n`, trailing `data: [DONE]\n\n`, OpenAI-compatible non-stream `choices`). `internal/agent/canvas/fixture_compile_test.go` + per-component tests pin the Python-equivalent outputs. ``` go test -count=1 -v -run 'TestRunAgent_RealCanvas|TestRunAgent_RunTracker' ./internal/service/ ``` ## Design reference `docs/develop/agent-go-port-design.md` (1329 lines, last cross-checked 2026-06-17) — module layout, per-component / per-tool inventory, corner-case catalogue, and the actionable backlog (Section 14, including the retrieval alignment follow-up). --------- Co-authored-by: Claude <noreply@anthropic.com>
2026-06-22 11:58:29 +08:00
req := SandboxRequest{
Lang: lang,
Script: script,
Arguments: args.Args,
Timeout: args.Timeout,
feat(agent): align Go agent behavior with Python (except retrieval component) (#16225) ## Summary Aligns the **Go agent runtime/canvas/components/tools** behavior with the **Python `agent/` implementation** so the same stored canvas DSL produces the same execution result on either side. Every component, tool, and runtime primitive in `internal/agent/` is now driven by the same semantics as its Python counterpart — variable resolution, template substitution, control flow, error reporting, retry/cancel, and stream event shapes. The **retrieval component is the one explicit exception** in this PR. It is being reworked in a separate change and is excluded from this alignment pass; the wrapper slot (`universe_a_wrappers.go → newRetrievalComponent`) is preserved. ## Scope of alignment ### Components (all aligned with `agent/component/`) `Begin` · `Message` · `LLM` (incl. ChatTemplateKwargs, MessageHistoryWindowSize, VisualFiles, Cite, OutputStructure, JSONOutput, TopP, MaxRetries, DelayAfterError, credentials) · `Agent` (react + tool artifact capture + `Reset()` interface-assert) · `Switch` (12/12 operators, Python-equivalent semantics) · `Categorize` · `Invoke` · `Iteration` · `Loop` (macro-expansion through `workflowx.AddLoopNode`) · `UserFillUp` (Python-equivalent interrupt/resume via eino `compose.Interrupt`/`ResumeWithData`) · `FillUp` · `DataOperations` · `ListOperations` · `StringTransform` · `VariableAggregator` · `VariableAssigner` · `Browser` (full stagehand runtime parity) · `DocsGenerator` · `ExcelProcessor`. ### Tools (all aligned with `agent/tools/`) `Retrieval` (wrapper slot only — logic out of scope) · `MCPToolAdapter` (streamable-HTTP) · `CodeExec` (sandbox bridge with `code_exec_contract.go` matching Python contract) · `AkShare` · `ArXiv` · `Crawler` · `DeepL` · `DuckDuckGo` · `Email` · `ExeSQL` · `GitHub` · `Google` · `GoogleScholar` · `Jin10` · `PubMed` · `QWeather` · `SearXNG` · `Tavily` · `Tushare` · `Wencai` · `Wikipedia` · `YahooFinance` — uniform `eino tool.InvokableTool` interface, SSRF protection, shared HTTP client. ### Canvas execution engine (`internal/agent/canvas/`) Aligned with Python's `agent/canvas.py`: - **Scheduler** (`scheduler.go`): state pre/post handlers, node lambdas, per-component timeout resolver (4-level: per-class env → per-class table → uniform env → 600s fallback), `legacyNoOpNames`. - **Loop subgraph** (`loop_subgraph.go`): Python-equivalent `AddLoopNode` macro expansion + condition translation. - **Multibranch** (`multibranch.go`): `Switch` / `Categorize` routing via `compose.NewGraphMultiBranch` — same branch selection semantics as Python. - **Parallel subgraph** (`parallel_subgraph.go`): matches Python's parallel fan-out contract. - **Interrupt/Resume** (`interrupt_resume.go`): `UserFillUpNodeBody` / `IsInterruptError` / `ExtractInterruptContexts` — replaces the deprecated Python sentinel chain with eino's native interrupt API, preserving the same external behavior. - **Checkpoint** (`checkpoint_store.go`): `RedisCheckPointStore` Get/Set/Delete, with business metadata (status / canvas_id / parent_run_id) on a parallel Redis Hash. - **RunTracker** (`run_tracker.go`): Start / MarkSucceeded / MarkFailed / MarkCancelled / AttachCheckpoint — same lifecycle as the Python run record. - **Cancel** (`cancel.go`): Redis pub/sub watch. - **Stream** (`stream.go`): SSE channel with `messages` / `waiting` / `errors` / `done` events, same shape as Python's `agent.canvas.RunEvent` payload. ### DSL bridge (`internal/agent/dsl/`) - `normalize.go`: v1↔v2 collapsed into a single wire format — Python and Go consume the same stored JSON. - `reset.go`: per-run state reset matches Python's `Canvas.reset()` semantics. - Testdata mirrors Python's `agent_msg.json` / `all.json` / etc. ### Runtime (`internal/agent/runtime/`) - `CanvasState` / `NewCanvasState` / `GetVar` / `SetVar` / `ReadVars`: same `{{cpn_id@param}}` resolution model. - `ResolveTemplate` (regex fast path + gonja fallback) — Python Jinja-style semantics. - `selector.go`, `metrics.go`, `component.go`: shared runtime contracts. ## Out of scope (intentionally) - **`Retrieval` component logic** — wrapped only; full parity lands in a follow-up PR. - **Frontend** — only minor dsl-bridge / canvas UX fixes ride along. - **CLI / admin / model registry** — orthogonal to agent behavior. ## How alignment is verified `internal/service/agent_run_e2e_test.go` exercises the **full production chain** against real Python-shaped DSL fixtures: ``` loadCanvasForUser → versionDAO.GetLatest → decodeCanvasFromDSL → canvas.Compile → cc.Workflow.Invoke → answer extraction ``` using in-memory SQLite + miniredis (no Docker). Covers: - `TestRunAgent_RealCanvas_BeginMessage` — happy path, `{{sys.query}}` resolution - `TestRunAgent_RealCanvas_WaitForUserResume` — two-run resume cycle (Python-equivalent) - `TestRunAgent_RealCanvas_CompileFails` — unknown component name → sanitized error (Python-equivalent) - `TestRunAgent_RealCanvas_InvokeFails` — unresolvable template ref (Python-equivalent) - `TestRunAgent_RunTracker_AttachCheckpoint_CallSequence` — Start→AttachCheckpoint→MarkSucceeded lifecycle `internal/handler/agent_test.go` — SSE streaming parity (`Content-Type: text/event-stream`, `data: {…}\n\n`, trailing `data: [DONE]\n\n`, OpenAI-compatible non-stream `choices`). `internal/agent/canvas/fixture_compile_test.go` + per-component tests pin the Python-equivalent outputs. ``` go test -count=1 -v -run 'TestRunAgent_RealCanvas|TestRunAgent_RunTracker' ./internal/service/ ``` ## Design reference `docs/develop/agent-go-port-design.md` (1329 lines, last cross-checked 2026-06-17) — module layout, per-component / per-tool inventory, corner-case catalogue, and the actionable backlog (Section 14, including the retrieval alignment follow-up). --------- Co-authored-by: Claude <noreply@anthropic.com>
2026-06-22 11:58:29 +08:00
}
common.Debug("CodeExec tool invoke",
zap.String("lang", req.Lang),
zap.Int("timeout", req.Timeout),
zap.Int("arguments_keys", len(req.Arguments)),
zap.Int("script_len", len(req.Script)))
feat(agent): align Go agent behavior with Python (except retrieval component) (#16225) ## Summary Aligns the **Go agent runtime/canvas/components/tools** behavior with the **Python `agent/` implementation** so the same stored canvas DSL produces the same execution result on either side. Every component, tool, and runtime primitive in `internal/agent/` is now driven by the same semantics as its Python counterpart — variable resolution, template substitution, control flow, error reporting, retry/cancel, and stream event shapes. The **retrieval component is the one explicit exception** in this PR. It is being reworked in a separate change and is excluded from this alignment pass; the wrapper slot (`universe_a_wrappers.go → newRetrievalComponent`) is preserved. ## Scope of alignment ### Components (all aligned with `agent/component/`) `Begin` · `Message` · `LLM` (incl. ChatTemplateKwargs, MessageHistoryWindowSize, VisualFiles, Cite, OutputStructure, JSONOutput, TopP, MaxRetries, DelayAfterError, credentials) · `Agent` (react + tool artifact capture + `Reset()` interface-assert) · `Switch` (12/12 operators, Python-equivalent semantics) · `Categorize` · `Invoke` · `Iteration` · `Loop` (macro-expansion through `workflowx.AddLoopNode`) · `UserFillUp` (Python-equivalent interrupt/resume via eino `compose.Interrupt`/`ResumeWithData`) · `FillUp` · `DataOperations` · `ListOperations` · `StringTransform` · `VariableAggregator` · `VariableAssigner` · `Browser` (full stagehand runtime parity) · `DocsGenerator` · `ExcelProcessor`. ### Tools (all aligned with `agent/tools/`) `Retrieval` (wrapper slot only — logic out of scope) · `MCPToolAdapter` (streamable-HTTP) · `CodeExec` (sandbox bridge with `code_exec_contract.go` matching Python contract) · `AkShare` · `ArXiv` · `Crawler` · `DeepL` · `DuckDuckGo` · `Email` · `ExeSQL` · `GitHub` · `Google` · `GoogleScholar` · `Jin10` · `PubMed` · `QWeather` · `SearXNG` · `Tavily` · `Tushare` · `Wencai` · `Wikipedia` · `YahooFinance` — uniform `eino tool.InvokableTool` interface, SSRF protection, shared HTTP client. ### Canvas execution engine (`internal/agent/canvas/`) Aligned with Python's `agent/canvas.py`: - **Scheduler** (`scheduler.go`): state pre/post handlers, node lambdas, per-component timeout resolver (4-level: per-class env → per-class table → uniform env → 600s fallback), `legacyNoOpNames`. - **Loop subgraph** (`loop_subgraph.go`): Python-equivalent `AddLoopNode` macro expansion + condition translation. - **Multibranch** (`multibranch.go`): `Switch` / `Categorize` routing via `compose.NewGraphMultiBranch` — same branch selection semantics as Python. - **Parallel subgraph** (`parallel_subgraph.go`): matches Python's parallel fan-out contract. - **Interrupt/Resume** (`interrupt_resume.go`): `UserFillUpNodeBody` / `IsInterruptError` / `ExtractInterruptContexts` — replaces the deprecated Python sentinel chain with eino's native interrupt API, preserving the same external behavior. - **Checkpoint** (`checkpoint_store.go`): `RedisCheckPointStore` Get/Set/Delete, with business metadata (status / canvas_id / parent_run_id) on a parallel Redis Hash. - **RunTracker** (`run_tracker.go`): Start / MarkSucceeded / MarkFailed / MarkCancelled / AttachCheckpoint — same lifecycle as the Python run record. - **Cancel** (`cancel.go`): Redis pub/sub watch. - **Stream** (`stream.go`): SSE channel with `messages` / `waiting` / `errors` / `done` events, same shape as Python's `agent.canvas.RunEvent` payload. ### DSL bridge (`internal/agent/dsl/`) - `normalize.go`: v1↔v2 collapsed into a single wire format — Python and Go consume the same stored JSON. - `reset.go`: per-run state reset matches Python's `Canvas.reset()` semantics. - Testdata mirrors Python's `agent_msg.json` / `all.json` / etc. ### Runtime (`internal/agent/runtime/`) - `CanvasState` / `NewCanvasState` / `GetVar` / `SetVar` / `ReadVars`: same `{{cpn_id@param}}` resolution model. - `ResolveTemplate` (regex fast path + gonja fallback) — Python Jinja-style semantics. - `selector.go`, `metrics.go`, `component.go`: shared runtime contracts. ## Out of scope (intentionally) - **`Retrieval` component logic** — wrapped only; full parity lands in a follow-up PR. - **Frontend** — only minor dsl-bridge / canvas UX fixes ride along. - **CLI / admin / model registry** — orthogonal to agent behavior. ## How alignment is verified `internal/service/agent_run_e2e_test.go` exercises the **full production chain** against real Python-shaped DSL fixtures: ``` loadCanvasForUser → versionDAO.GetLatest → decodeCanvasFromDSL → canvas.Compile → cc.Workflow.Invoke → answer extraction ``` using in-memory SQLite + miniredis (no Docker). Covers: - `TestRunAgent_RealCanvas_BeginMessage` — happy path, `{{sys.query}}` resolution - `TestRunAgent_RealCanvas_WaitForUserResume` — two-run resume cycle (Python-equivalent) - `TestRunAgent_RealCanvas_CompileFails` — unknown component name → sanitized error (Python-equivalent) - `TestRunAgent_RealCanvas_InvokeFails` — unresolvable template ref (Python-equivalent) - `TestRunAgent_RunTracker_AttachCheckpoint_CallSequence` — Start→AttachCheckpoint→MarkSucceeded lifecycle `internal/handler/agent_test.go` — SSE streaming parity (`Content-Type: text/event-stream`, `data: {…}\n\n`, trailing `data: [DONE]\n\n`, OpenAI-compatible non-stream `choices`). `internal/agent/canvas/fixture_compile_test.go` + per-component tests pin the Python-equivalent outputs. ``` go test -count=1 -v -run 'TestRunAgent_RealCanvas|TestRunAgent_RunTracker' ./internal/service/ ``` ## Design reference `docs/develop/agent-go-port-design.md` (1329 lines, last cross-checked 2026-06-17) — module layout, per-component / per-tool inventory, corner-case catalogue, and the actionable backlog (Section 14, including the retrieval alignment follow-up). --------- Co-authored-by: Claude <noreply@anthropic.com>
2026-06-22 11:58:29 +08:00
resp, err := client.ExecuteCode(ctx, req)
if err != nil {
return codeExecStubResult(err.Error()), err
}
out, mErr := codeExecResultJSON(resp)
if mErr != nil {
return codeExecStubResult(mErr.Error()), mErr
}
return out, nil
}
// codeExecResultJSON serializes a SandboxResponse into the envelope
// the eino tool contract returns. Field mapping mirrors the Python
// tool's `code_exec.py:385-490` `_process_execution_result`:
//
// - Stdout / Stderr / ExitCode: stream directly through.
// - Returned → Content (the model's natural "what did main() give
// us back" field).
// - StructuredResult["actual_type"] → ActualType (Python
// `infer_actual_type` surface for downstream Message component).
// - Metadata["artifacts"] → Artifacts (the model AND the Message
// component's `_ARTIFACTS` collector both consume this; we
// surface it as `_ARTIFACTS` to match the Python envelope).
// - Metadata["attachments"] → Attachments (rendered into
// downstream Markdown by Message via the same path the Agent
// tool artifact markdown uses).
//
// Artifacts / Attachments with the wrong element type (anything
// other than map[string]any) are silently dropped with a log
// warning. This matches the Python tool's "skip on shape mismatch"
// semantics — better to lose one artifact than to abort the run.
func codeExecResultJSON(r *SandboxResponse) (string, error) {
if r == nil {
return codeExecStubResult("empty response"), nil
}
out := codeExecResult{
ExitCode: r.ExitCode,
Stdout: r.Stdout,
Stderr: r.Stderr,
}
if r.Metadata != nil {
out.Artifacts = extractArtifactList(r.Metadata, "artifacts")
out.Attachments = extractArtifactList(r.Metadata, "attachments")
}
feat(agent): align Go agent behavior with Python (except retrieval component) (#16225) ## Summary Aligns the **Go agent runtime/canvas/components/tools** behavior with the **Python `agent/` implementation** so the same stored canvas DSL produces the same execution result on either side. Every component, tool, and runtime primitive in `internal/agent/` is now driven by the same semantics as its Python counterpart — variable resolution, template substitution, control flow, error reporting, retry/cancel, and stream event shapes. The **retrieval component is the one explicit exception** in this PR. It is being reworked in a separate change and is excluded from this alignment pass; the wrapper slot (`universe_a_wrappers.go → newRetrievalComponent`) is preserved. ## Scope of alignment ### Components (all aligned with `agent/component/`) `Begin` · `Message` · `LLM` (incl. ChatTemplateKwargs, MessageHistoryWindowSize, VisualFiles, Cite, OutputStructure, JSONOutput, TopP, MaxRetries, DelayAfterError, credentials) · `Agent` (react + tool artifact capture + `Reset()` interface-assert) · `Switch` (12/12 operators, Python-equivalent semantics) · `Categorize` · `Invoke` · `Iteration` · `Loop` (macro-expansion through `workflowx.AddLoopNode`) · `UserFillUp` (Python-equivalent interrupt/resume via eino `compose.Interrupt`/`ResumeWithData`) · `FillUp` · `DataOperations` · `ListOperations` · `StringTransform` · `VariableAggregator` · `VariableAssigner` · `Browser` (full stagehand runtime parity) · `DocsGenerator` · `ExcelProcessor`. ### Tools (all aligned with `agent/tools/`) `Retrieval` (wrapper slot only — logic out of scope) · `MCPToolAdapter` (streamable-HTTP) · `CodeExec` (sandbox bridge with `code_exec_contract.go` matching Python contract) · `AkShare` · `ArXiv` · `Crawler` · `DeepL` · `DuckDuckGo` · `Email` · `ExeSQL` · `GitHub` · `Google` · `GoogleScholar` · `Jin10` · `PubMed` · `QWeather` · `SearXNG` · `Tavily` · `Tushare` · `Wencai` · `Wikipedia` · `YahooFinance` — uniform `eino tool.InvokableTool` interface, SSRF protection, shared HTTP client. ### Canvas execution engine (`internal/agent/canvas/`) Aligned with Python's `agent/canvas.py`: - **Scheduler** (`scheduler.go`): state pre/post handlers, node lambdas, per-component timeout resolver (4-level: per-class env → per-class table → uniform env → 600s fallback), `legacyNoOpNames`. - **Loop subgraph** (`loop_subgraph.go`): Python-equivalent `AddLoopNode` macro expansion + condition translation. - **Multibranch** (`multibranch.go`): `Switch` / `Categorize` routing via `compose.NewGraphMultiBranch` — same branch selection semantics as Python. - **Parallel subgraph** (`parallel_subgraph.go`): matches Python's parallel fan-out contract. - **Interrupt/Resume** (`interrupt_resume.go`): `UserFillUpNodeBody` / `IsInterruptError` / `ExtractInterruptContexts` — replaces the deprecated Python sentinel chain with eino's native interrupt API, preserving the same external behavior. - **Checkpoint** (`checkpoint_store.go`): `RedisCheckPointStore` Get/Set/Delete, with business metadata (status / canvas_id / parent_run_id) on a parallel Redis Hash. - **RunTracker** (`run_tracker.go`): Start / MarkSucceeded / MarkFailed / MarkCancelled / AttachCheckpoint — same lifecycle as the Python run record. - **Cancel** (`cancel.go`): Redis pub/sub watch. - **Stream** (`stream.go`): SSE channel with `messages` / `waiting` / `errors` / `done` events, same shape as Python's `agent.canvas.RunEvent` payload. ### DSL bridge (`internal/agent/dsl/`) - `normalize.go`: v1↔v2 collapsed into a single wire format — Python and Go consume the same stored JSON. - `reset.go`: per-run state reset matches Python's `Canvas.reset()` semantics. - Testdata mirrors Python's `agent_msg.json` / `all.json` / etc. ### Runtime (`internal/agent/runtime/`) - `CanvasState` / `NewCanvasState` / `GetVar` / `SetVar` / `ReadVars`: same `{{cpn_id@param}}` resolution model. - `ResolveTemplate` (regex fast path + gonja fallback) — Python Jinja-style semantics. - `selector.go`, `metrics.go`, `component.go`: shared runtime contracts. ## Out of scope (intentionally) - **`Retrieval` component logic** — wrapped only; full parity lands in a follow-up PR. - **Frontend** — only minor dsl-bridge / canvas UX fixes ride along. - **CLI / admin / model registry** — orthogonal to agent behavior. ## How alignment is verified `internal/service/agent_run_e2e_test.go` exercises the **full production chain** against real Python-shaped DSL fixtures: ``` loadCanvasForUser → versionDAO.GetLatest → decodeCanvasFromDSL → canvas.Compile → cc.Workflow.Invoke → answer extraction ``` using in-memory SQLite + miniredis (no Docker). Covers: - `TestRunAgent_RealCanvas_BeginMessage` — happy path, `{{sys.query}}` resolution - `TestRunAgent_RealCanvas_WaitForUserResume` — two-run resume cycle (Python-equivalent) - `TestRunAgent_RealCanvas_CompileFails` — unknown component name → sanitized error (Python-equivalent) - `TestRunAgent_RealCanvas_InvokeFails` — unresolvable template ref (Python-equivalent) - `TestRunAgent_RunTracker_AttachCheckpoint_CallSequence` — Start→AttachCheckpoint→MarkSucceeded lifecycle `internal/handler/agent_test.go` — SSE streaming parity (`Content-Type: text/event-stream`, `data: {…}\n\n`, trailing `data: [DONE]\n\n`, OpenAI-compatible non-stream `choices`). `internal/agent/canvas/fixture_compile_test.go` + per-component tests pin the Python-equivalent outputs. ``` go test -count=1 -v -run 'TestRunAgent_RealCanvas|TestRunAgent_RunTracker' ./internal/service/ ``` ## Design reference `docs/develop/agent-go-port-design.md` (1329 lines, last cross-checked 2026-06-17) — module layout, per-component / per-tool inventory, corner-case catalogue, and the actionable backlog (Section 14, including the retrieval alignment follow-up). --------- Co-authored-by: Claude <noreply@anthropic.com>
2026-06-22 11:58:29 +08:00
hasStructuredResult := false
resolvedValue, usedStdoutFallback := resolveCodeExecResultValue(r)
if r.StructuredResult != nil {
hasStructuredResult, _ = r.StructuredResult["present"].(bool)
}
if strings.TrimSpace(r.Stderr) != "" &&
!hasStructuredResult &&
len(out.Artifacts) == 0 &&
strings.TrimSpace(r.Stdout) == "" {
out.Error = r.Stderr
} else {
if usedStdoutFallback && strings.TrimSpace(r.Stdout) != "" {
fmt.Fprintln(os.Stderr, "code_exec: falling back to stdout deserialization because no structured result metadata was provided")
}
out.RawResult = NormalizeCodeExecOutputValue(resolvedValue)
out.ActualType = InferCodeExecActualType(out.RawResult)
out.Content = RenderCodeExecCanonicalContent(out.RawResult)
}
common.Debug("CodeExec tool",
zap.Any("structured_result", r.StructuredResult),
zap.Any("resolved_value", resolvedValue),
zap.Any("raw_result", out.RawResult),
zap.String("content", out.Content),
zap.String("actual_type", out.ActualType),
zap.Bool("stderr_present", r.Stderr != ""),
zap.Int("stderr_len", len(r.Stderr)),
zap.Int("stdout_len", len(r.Stdout)))
b, err := json.Marshal(out)
if err != nil {
return "", fmt.Errorf("code_exec: marshal result: %w", err)
}
return string(b), nil
}
// extractArtifactList pulls a list of dict-shaped entries out of
// Metadata[key]. Items that aren't map[string]any are dropped with
// a stderr log line so the operator can see the data loss without
// the run aborting.
func extractArtifactList(meta map[string]any, key string) []map[string]any {
raw, ok := meta[key]
if !ok {
return nil
}
arr, ok := raw.([]any)
if !ok {
return nil
}
out := make([]map[string]any, 0, len(arr))
for i, item := range arr {
m, ok := item.(map[string]any)
if !ok {
fmt.Fprintf(os.Stderr, "code_exec: %s[%d] is %T, expected map[string]any; dropping\n", key, i, item)
continue
}
out = append(out, m)
}
return out
feat(agent): Go port — canvas engine, 22 components, DSL v2, 13 endpoints (#15952) Ports the agent canvas subsystem from Python to Go. ## What's included ### Canvas Engine (Phase 0/1) - State engine, scheduler, variable resolver, Redis checkpoint store, cancel protocol - **209 tests** across canvas / component / io packages ### 22 Components (P0–P4) | Tier | Components | |---|---| | P0 T1+T2+T3 | LLM, Agent, ExitLoop, Switch, Categorize, Begin, Message, Invoke | | P1 T3 | VariableAggregator, VariableAssigner, StringTransform, ListOperations, DataOperations | | P2 T3 | Iteration, IterationItem, Loop, LoopItem | | P3 T3 | UserFillUp, Fillup | | P4 T5 | Browser, ExcelProcessor, DocsGenerator | ### DSL v2 Schema (Phase 2.5) - Typed v2 in-memory model with v1-to-v2 auto-detect converter - v1 legacy field stripping per plan §2.11.7 ### HTTP Endpoints & Bug Fixes (Plans PR1–PR3) - **DELETE SQL bug fix**: gorm v2 `Where("id = ?", id).Delete(...)` pattern - **CreateAgent validation**: title/DSL required, duplicate check, 103 envelope - **13 new endpoints**: templates, prompts, tags, sessions CRUD, chat/completions (SSE + non-stream stubs), rerun, test_db_connection, logs, webhook/logs - **756 Go unit tests** (745 → 756, +18) - **17 → 0 Python integration test failures** (test_agents.py + test_session_management/) ### Tools 21 eino tools: HTTPHelper, search tools, financial/data tools, mandatory stubs ### Infrastructure OTel observability, NATS message queue, DeepDoc gRPC client, SSRF guards, IDOR mitigation
2026-06-12 22:58:28 +08:00
}
func codeExecStubResult(msg string) string {
b, err := json.Marshal(codeExecResult{
Stub: true,
Error: msg,
})
if err != nil {
return fmt.Sprintf(`{"_ERROR":"code_exec: marshal stub: %s","stub":true}`, err)
}
return string(b)
}
feat(agent): align Go agent behavior with Python (except retrieval component) (#16225) ## Summary Aligns the **Go agent runtime/canvas/components/tools** behavior with the **Python `agent/` implementation** so the same stored canvas DSL produces the same execution result on either side. Every component, tool, and runtime primitive in `internal/agent/` is now driven by the same semantics as its Python counterpart — variable resolution, template substitution, control flow, error reporting, retry/cancel, and stream event shapes. The **retrieval component is the one explicit exception** in this PR. It is being reworked in a separate change and is excluded from this alignment pass; the wrapper slot (`universe_a_wrappers.go → newRetrievalComponent`) is preserved. ## Scope of alignment ### Components (all aligned with `agent/component/`) `Begin` · `Message` · `LLM` (incl. ChatTemplateKwargs, MessageHistoryWindowSize, VisualFiles, Cite, OutputStructure, JSONOutput, TopP, MaxRetries, DelayAfterError, credentials) · `Agent` (react + tool artifact capture + `Reset()` interface-assert) · `Switch` (12/12 operators, Python-equivalent semantics) · `Categorize` · `Invoke` · `Iteration` · `Loop` (macro-expansion through `workflowx.AddLoopNode`) · `UserFillUp` (Python-equivalent interrupt/resume via eino `compose.Interrupt`/`ResumeWithData`) · `FillUp` · `DataOperations` · `ListOperations` · `StringTransform` · `VariableAggregator` · `VariableAssigner` · `Browser` (full stagehand runtime parity) · `DocsGenerator` · `ExcelProcessor`. ### Tools (all aligned with `agent/tools/`) `Retrieval` (wrapper slot only — logic out of scope) · `MCPToolAdapter` (streamable-HTTP) · `CodeExec` (sandbox bridge with `code_exec_contract.go` matching Python contract) · `AkShare` · `ArXiv` · `Crawler` · `DeepL` · `DuckDuckGo` · `Email` · `ExeSQL` · `GitHub` · `Google` · `GoogleScholar` · `Jin10` · `PubMed` · `QWeather` · `SearXNG` · `Tavily` · `Tushare` · `Wencai` · `Wikipedia` · `YahooFinance` — uniform `eino tool.InvokableTool` interface, SSRF protection, shared HTTP client. ### Canvas execution engine (`internal/agent/canvas/`) Aligned with Python's `agent/canvas.py`: - **Scheduler** (`scheduler.go`): state pre/post handlers, node lambdas, per-component timeout resolver (4-level: per-class env → per-class table → uniform env → 600s fallback), `legacyNoOpNames`. - **Loop subgraph** (`loop_subgraph.go`): Python-equivalent `AddLoopNode` macro expansion + condition translation. - **Multibranch** (`multibranch.go`): `Switch` / `Categorize` routing via `compose.NewGraphMultiBranch` — same branch selection semantics as Python. - **Parallel subgraph** (`parallel_subgraph.go`): matches Python's parallel fan-out contract. - **Interrupt/Resume** (`interrupt_resume.go`): `UserFillUpNodeBody` / `IsInterruptError` / `ExtractInterruptContexts` — replaces the deprecated Python sentinel chain with eino's native interrupt API, preserving the same external behavior. - **Checkpoint** (`checkpoint_store.go`): `RedisCheckPointStore` Get/Set/Delete, with business metadata (status / canvas_id / parent_run_id) on a parallel Redis Hash. - **RunTracker** (`run_tracker.go`): Start / MarkSucceeded / MarkFailed / MarkCancelled / AttachCheckpoint — same lifecycle as the Python run record. - **Cancel** (`cancel.go`): Redis pub/sub watch. - **Stream** (`stream.go`): SSE channel with `messages` / `waiting` / `errors` / `done` events, same shape as Python's `agent.canvas.RunEvent` payload. ### DSL bridge (`internal/agent/dsl/`) - `normalize.go`: v1↔v2 collapsed into a single wire format — Python and Go consume the same stored JSON. - `reset.go`: per-run state reset matches Python's `Canvas.reset()` semantics. - Testdata mirrors Python's `agent_msg.json` / `all.json` / etc. ### Runtime (`internal/agent/runtime/`) - `CanvasState` / `NewCanvasState` / `GetVar` / `SetVar` / `ReadVars`: same `{{cpn_id@param}}` resolution model. - `ResolveTemplate` (regex fast path + gonja fallback) — Python Jinja-style semantics. - `selector.go`, `metrics.go`, `component.go`: shared runtime contracts. ## Out of scope (intentionally) - **`Retrieval` component logic** — wrapped only; full parity lands in a follow-up PR. - **Frontend** — only minor dsl-bridge / canvas UX fixes ride along. - **CLI / admin / model registry** — orthogonal to agent behavior. ## How alignment is verified `internal/service/agent_run_e2e_test.go` exercises the **full production chain** against real Python-shaped DSL fixtures: ``` loadCanvasForUser → versionDAO.GetLatest → decodeCanvasFromDSL → canvas.Compile → cc.Workflow.Invoke → answer extraction ``` using in-memory SQLite + miniredis (no Docker). Covers: - `TestRunAgent_RealCanvas_BeginMessage` — happy path, `{{sys.query}}` resolution - `TestRunAgent_RealCanvas_WaitForUserResume` — two-run resume cycle (Python-equivalent) - `TestRunAgent_RealCanvas_CompileFails` — unknown component name → sanitized error (Python-equivalent) - `TestRunAgent_RealCanvas_InvokeFails` — unresolvable template ref (Python-equivalent) - `TestRunAgent_RunTracker_AttachCheckpoint_CallSequence` — Start→AttachCheckpoint→MarkSucceeded lifecycle `internal/handler/agent_test.go` — SSE streaming parity (`Content-Type: text/event-stream`, `data: {…}\n\n`, trailing `data: [DONE]\n\n`, OpenAI-compatible non-stream `choices`). `internal/agent/canvas/fixture_compile_test.go` + per-component tests pin the Python-equivalent outputs. ``` go test -count=1 -v -run 'TestRunAgent_RealCanvas|TestRunAgent_RunTracker' ./internal/service/ ``` ## Design reference `docs/develop/agent-go-port-design.md` (1329 lines, last cross-checked 2026-06-17) — module layout, per-component / per-tool inventory, corner-case catalogue, and the actionable backlog (Section 14, including the retrieval alignment follow-up). --------- Co-authored-by: Claude <noreply@anthropic.com>
2026-06-22 11:58:29 +08:00
func resolveCodeExecResultValue(r *SandboxResponse) (any, bool) {
if r != nil && r.StructuredResult != nil {
if present, _ := r.StructuredResult["present"].(bool); present {
return r.StructuredResult["value"], false
}
}
return deserializeCodeExecStdout(r.Stdout), true
}
func deserializeCodeExecStdout(stdout string) any {
text := strings.TrimSpace(stdout)
if text == "" {
return ""
}
var decoded any
if err := json.Unmarshal([]byte(text), &decoded); err == nil {
return decoded
}
return text
}
feat(agent): Go port — canvas engine, 22 components, DSL v2, 13 endpoints (#15952) Ports the agent canvas subsystem from Python to Go. ## What's included ### Canvas Engine (Phase 0/1) - State engine, scheduler, variable resolver, Redis checkpoint store, cancel protocol - **209 tests** across canvas / component / io packages ### 22 Components (P0–P4) | Tier | Components | |---|---| | P0 T1+T2+T3 | LLM, Agent, ExitLoop, Switch, Categorize, Begin, Message, Invoke | | P1 T3 | VariableAggregator, VariableAssigner, StringTransform, ListOperations, DataOperations | | P2 T3 | Iteration, IterationItem, Loop, LoopItem | | P3 T3 | UserFillUp, Fillup | | P4 T5 | Browser, ExcelProcessor, DocsGenerator | ### DSL v2 Schema (Phase 2.5) - Typed v2 in-memory model with v1-to-v2 auto-detect converter - v1 legacy field stripping per plan §2.11.7 ### HTTP Endpoints & Bug Fixes (Plans PR1–PR3) - **DELETE SQL bug fix**: gorm v2 `Where("id = ?", id).Delete(...)` pattern - **CreateAgent validation**: title/DSL required, duplicate check, 103 envelope - **13 new endpoints**: templates, prompts, tags, sessions CRUD, chat/completions (SSE + non-stream stubs), rerun, test_db_connection, logs, webhook/logs - **756 Go unit tests** (745 → 756, +18) - **17 → 0 Python integration test failures** (test_agents.py + test_session_management/) ### Tools 21 eino tools: HTTPHelper, search tools, financial/data tools, mandatory stubs ### Infrastructure OTel observability, NATS message queue, DeepDoc gRPC client, SSRF guards, IDOR mitigation
2026-06-12 22:58:28 +08:00
// normalizeCodeExecLang accepts the model's literal "language" or the
// Python-style "lang" alias and maps synonyms to the canonical "python" /
// "nodejs" forms used by the Python sandbox.
func normalizeCodeExecLang(primary, alias string) string {
v := strings.ToLower(strings.TrimSpace(primary))
if v == "" {
v = strings.ToLower(strings.TrimSpace(alias))
}
switch v {
case "python", "python3":
return "python"
case "javascript", "js", "nodejs", "node":
return "nodejs"
}
return ""
}