Files
ragflow/internal/entity/models/base_model.go

352 lines
10 KiB
Go
Raw Normal View History

//
// 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 models
import (
"bufio"
"bytes"
"context"
"encoding/json"
"fmt"
"io"
"net/http"
"sort"
"strings"
"time"
)
type BaseModel struct {
BaseURL map[string]string
URLSuffix URLSuffix
httpClient *http.Client
AllowEmptyAPIKey bool
}
func (b *BaseModel) APIConfigCheck(apiConfig *APIConfig) error {
if b.AllowEmptyAPIKey {
return nil
}
if apiConfig == nil || apiConfig.ApiKey == nil || strings.TrimSpace(*apiConfig.ApiKey) == "" {
return fmt.Errorf("api key is required")
}
return nil
}
// BearerAuth returns the Bearer token for Authorization header,
// or empty string if apiConfig or its ApiKey is nil/empty.
func BearerAuth(apiConfig *APIConfig) string {
if apiConfig == nil || apiConfig.ApiKey == nil {
return ""
}
key := strings.TrimSpace(*apiConfig.ApiKey)
if key == "" {
return ""
}
return fmt.Sprintf("Bearer %s", key)
}
func (b *BaseModel) GetBaseURL(apiConfig *APIConfig) (string, error) {
if apiConfig != nil && apiConfig.BaseURL != nil && *apiConfig.BaseURL != "" {
return strings.TrimSuffix(*apiConfig.BaseURL, "/"), nil
}
region := "default"
hasRegion := false
if apiConfig != nil && apiConfig.Region != nil {
hasRegion = true
region = *apiConfig.Region
}
baseURL, ok := b.BaseURL[region]
if !ok || baseURL == "" {
if (!hasRegion || region == "") && b.BaseURL != nil {
if defaultBaseURL, ok := b.BaseURL["default"]; ok && defaultBaseURL != "" {
return defaultBaseURL, nil
}
}
return "", fmt.Errorf("no base URL configured for region %q", region)
}
baseURL = strings.TrimSuffix(baseURL, "/")
return baseURL, nil
}
// ParseSSEStream reads the body of an OpenAI-compatible Server-Sent Events
// response and calls onEvent for each successfully-parsed JSON payload.
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
// A malformed JSON payload after "data:" returns an error wrapped as
// "invalid SSE event" so the caller cannot silently swallow truncated or
// corrupted streams.
func ParseSSEStream[T any](r io.Reader, onEvent func(event T) error) (done bool, err error) {
scanner := bufio.NewScanner(r)
scanner.Buffer(make([]byte, 64*1024), 1024*1024)
for scanner.Scan() {
line := scanner.Text()
if !strings.HasPrefix(line, "data:") {
continue
}
data := strings.TrimSpace(line[5:])
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
if data == "" {
continue
}
if data == "[DONE]" {
return true, nil
}
var event T
if err := json.Unmarshal([]byte(data), &event); err != nil {
return false, fmt.Errorf("invalid SSE event: %w", err)
}
if err := onEvent(event); err != nil {
return false, err
}
}
return false, scanner.Err()
}
// ParseSSEStreamTolerant is like ParseSSEStream but silently skips
// malformed JSON payloads. Use this only for drivers whose upstream is
// known to interleave invalid frames the test suite documents as safe
// to ignore.
func ParseSSEStreamTolerant[T any](r io.Reader, onEvent func(event T) error) (done bool, err error) {
scanner := bufio.NewScanner(r)
scanner.Buffer(make([]byte, 64*1024), 1024*1024)
for scanner.Scan() {
line := scanner.Text()
if !strings.HasPrefix(line, "data:") {
continue
}
data := strings.TrimSpace(line[5:])
if data == "" {
continue
}
if data == "[DONE]" {
return true, nil
}
var event T
if err := json.Unmarshal([]byte(data), &event); err != nil {
continue
}
if err := onEvent(event); err != nil {
return false, err
}
}
return false, scanner.Err()
}
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
// ParseListModel Parse model list. Empty/whitespace IDs are skipped so
// upstream typos do not surface as blank entries in the UI.
func ParseListModel(modelList ModelList) []ListModelResponse {
var models []ListModelResponse
pm := GetProviderManager()
for _, model := range modelList.Models {
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
modelName := strings.TrimSpace(model.ID)
if modelName == "" {
continue
}
var modelResponse ListModelResponse
var modelEntity *Model
if pm != nil {
modelEntity = pm.GetModelByNameOrAlias(modelName)
}
if model.OwnedBy != "" {
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
modelName = modelName + "@" + model.OwnedBy
}
modelResponse.Name = modelName
if modelEntity != nil {
modelResponse.MaxDimension = modelEntity.MaxDimension
modelResponse.Dimensions = modelEntity.Dimensions
modelResponse.MaxTokens = modelEntity.MaxTokens
modelResponse.ModelTypes = modelEntity.ModelTypes
modelResponse.Thinking = modelEntity.Thinking
modelResponse.Dimensions = modelEntity.Dimensions
}
models = append(models, modelResponse)
}
return models
}
// NewDriverHTTPClient returns an *http.Client with the standard connection-pool
func NewDriverHTTPClient() *http.Client {
var t *http.Transport
if dt, ok := http.DefaultTransport.(*http.Transport); ok {
t = dt.Clone()
} else {
t = &http.Transport{Proxy: http.ProxyFromEnvironment}
}
t.MaxIdleConns = 100
t.MaxIdleConnsPerHost = 10
t.IdleConnTimeout = 90 * time.Second
t.DisableCompression = false
fix(ingestion): align laws DSL with Python — heading fallback, colon-title, short-line filter, remove_toc, and image extension mapping (#17200) ## Summary This PR aligns the Go ingestion pipeline's **Laws** DSL template with the Python implementation by fixing heading-detection gaps, adds image-extension support, refactors the **Extractor** component's LLM resolution, hardens heading detection for CJK text, and makes the Extractor accept the Python DSL prompt key names (`sys_prompt`/`prompts`) alongside the Go names. ## Changes ### 1. Picture file-type detection (`internal/utility/file.go`) Adds explicit mapping for common image extensions (png, jpg, jpeg, gif, bmp, tiff, tif, webp, svg, ico, avif, heic, apng) → `FileTypeVISUAL`, with regression tests. ### 2. Laws DSL heading-detection alignment (`internal/ingestion/component/chunker/`) Four fixes to `resolveTitleLevels`: | Fix | What changed | Why | |-----|-------------|-----| | **DOCX `ck_type` fallback** | `ckType` field on `lineRecord`, propagated from `ChunkDoc.CKType` in `recordsFromStructured`. When `ck_type=="heading"`, assign `fallbackLevel`. | office_oxide extracts DOCX heading metadata, but the info was lost before reaching the heading detector. Word headings whose text doesn't match any regex (e.g. "Introduction") were treated as body. | | **`make_colon_as_title` promotion** | `isColonTitle()`: promotes lines ending with `:`/`:` that have sentence-ending punctuation before the colon and ≥32 runes between them. | Mirrors Python's `make_colon_as_title` in `rag/nlp/__init__.py`. Triple guard prevents false positives. | | **Short/numeric line filter** | Lines with ≤1 rune or purely numeric are pinned to body level. | Mirrors Python `tree_merge`'s filter of `sections` where `len(...) <= 1` or `re.match(r"[0-9]+$", ...)`. | | **PDF `remove_toc`** | `"remove_toc": true` added to the PDF parser setup in `ingestion_pipeline_laws.json`. | The Go PDF parser already supports TOC removal; the Book template already enables it. | ### 3. Extractor llm_id resolution (`internal/ingestion/component/extractor.go`) Refactored to handle both **bare tenant_model UUIDs** and **composite model@provider** strings via the shared `resolveModelConfig` (`dispatch_model.go`): - **`resolveExtractorChatConfig`** — UUID path calls `resolveModelConfigByID` directly (one DB hit); composite path goes through `resolveModelConfig`. Added `isBareTenantModelID` pre-check for clear errors when a UUID doesn't exist. - **`resolveExtractorChatTarget`** — propagates resolution errors instead of silently returning empty driver. - **`Chat()`** — removed `driver = "dummy"` fallback. Missing driver is now an explicit error. - **Removed dead code**: `splitExtractorLLID`, `findExtractorSoleActiveInstance`. ### 4. `InjectExtractorLLMID` — fallback when no user config (`internal/common/parser_config.go`) Injects the tenant's global default LLM into extractor components **only when their `llm_id` is empty**. Preserves user-selected UUID or model@provider values. Priority: user-configured llm_id > tenant global default > error (no silent dummy fallback). ### 5. `ResponseHeaderTimeout` increase (`internal/entity/models/base_model.go`) `ResponseHeaderTimeout` 60s → 120s in `NewDriverHTTPClient`. Reasoning models with large extraction prompts can take longer than 60s to produce the first response token. ### 6. CJK rune-aware heading detection (`internal/ingestion/component/chunker/title.go`) Two byte-vs-rune bugs that only manifest on CJK text: | Fix | What changed | Why | |-----|-------------|-----| | **`isColonTitle` byte offset** | `body[lastPunct+1:]` → `body[lastPunct+runeLen:]` via `utf8.DecodeRuneInString` | `strings.LastIndexAny` returns a byte index; `+1` skips only 1 byte, corrupting multi-byte CJK punctuation (e.g. `。` = 3 bytes) and inflating the rune count past the 32-rune threshold → false-positive heading promotion. | | **Short-line filter byte count** | `len(text) <= 1` → `utf8.RuneCountInString(text) <= 1` | Go `len` is UTF-8 bytes; a single CJK char (3 bytes) passed the filter, but Python's `len` returns 1 → mismatch. | ### 7. `extractor_tag.go` — log error when llm fails When `resolveExtractorChatTarget` returned an error, `runAutoTags` will log error. ### 8. Python DSL prompt-key compatibility (`internal/ingestion/component/extractor.go`) The Resume DSL template uses Python-side key names (`sys_prompt`, `prompts`). `NewExtractorComponent` now accepts them as fallbacks alongside the Go names: - `system_prompt` (Go) ← `sys_prompt` (Python) as fallback - `prompt` (Go string) ← `prompts` (Python array `[{"role","content"}]`, takes `[0].content`) as fallback Mirrors the alias pattern already in `internal/agent/component/llm.go`. `resolveInputs` accepts per-call `sys_prompt` override too. ## Remaining gaps vs Python | Gap | Scope | Impact | |-----|-------|--------| | **TOC removal for TXT/MD/HTML** | Python's `remove_contents_table` works on all text formats; Go's `remove_toc` is PDF-only. | Low — plain-text documents rarely contain structured TOCs. | | **Regex pattern details** | Minor differences in quantifiers, missing H5/H6 markdown patterns, missing 4-level numbering pattern. | Low — Go's variants are stricter; DOCX headings are covered by `ck_type` fallback. | ## Testing - `TestHierarchyTitleChunker_CKTypeHeadingFallback` — DOCX `ck_type` heading promotion - `TestHierarchyTitleChunker_ColonTitlePromotion` / `_ColonTitleShortLine_Negative` — colon-title promotion + guard - `TestIsColonTitle_CJKEdgeCase` / `TestIsColonTitle_ASCII_NoRegression` — CJK byte-offset fix + ASCII regression - `TestHierarchyTitleChunker_ColonTitlePromotion_CJK_EdgeCase` — CJK colon edge case through full pipeline - `TestHierarchyTitleChunker_ShortSingleCJKLineFilter` — single CJK char filtered to body - `TestHierarchyTitleChunker_ShortNumericLineFilter` — purely numeric lines filtered - `TestGetFileType_ImageExtensions` / `_ExistingFormats_NoRegression` — image extension mapping - `TestInjectExtractorLLMID_SkipWhenUUID` / `_SkipWhenComposite` / `_InjectWhenEmpty` — llm_id injection guard - `TestIsBareTenantModelID` — UUID detection - `TestResolveExtractorChatTarget_AtSplitFallback` / `_NoDriver` — @ split fallback without DB - `TestNewExtractorComponent_SysPromptAlias` / `_PromptsArray` / `_PromptsArray_PromptWins` / `_SystemPromptWinsOverSysPrompt` — Python key compatibility - `TestBuildDOCXJSONSections_List` / `_TextBox` / `_MixedWithList` — DOCX list/text_box parsing - Full ingestion test suite passes (chunker, pipeline, task, service, component packages)
2026-07-22 19:14:32 +08:00
t.ResponseHeaderTimeout = 2 * 60 * time.Second
t.TLSHandshakeTimeout = 30 * time.Second
return &http.Client{Transport: t}
}
// PostJSONRequest marshals body to JSON, creates a POST request to url
func PostJSONRequest(ctx context.Context, client *http.Client, url, auth string, body map[string]interface{}) (*http.Response, error) {
data, err := json.Marshal(body)
if err != nil {
return nil, fmt.Errorf("failed to marshal request: %w", err)
}
req, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewBuffer(data))
if err != nil {
return nil, fmt.Errorf("failed to create request: %w", err)
}
req.Header.Set("Content-Type", "application/json")
if auth != "" {
req.Header.Set("Authorization", auth)
}
return client.Do(req)
}
// ReadErrorBody reads all bytes from r and returns them as a string suitable
func ReadErrorBody(r io.Reader) string {
b, _ := io.ReadAll(r)
return string(b)
}
// buildChatMessages converts internal messages to chat API payload items.
func buildChatMessages(messages []Message) []map[string]any {
apiMessages := make([]map[string]interface{}, len(messages))
for i, msg := range messages {
apiMsg := map[string]interface{}{
"role": msg.Role,
"content": msg.Content,
}
if msg.ToolCallID != "" {
apiMsg["tool_call_id"] = msg.ToolCallID
}
if len(msg.ToolCalls) > 0 {
apiMsg["tool_calls"] = msg.ToolCalls
}
apiMessages[i] = apiMsg
}
return apiMessages
}
// applyChatToolConfig adds OpenAI-compatible tool configuration to a request.
func applyChatToolConfig(reqBody map[string]interface{}, chatConfig *ChatConfig) {
if chatConfig == nil || chatConfig.Tools == nil {
return
}
reqBody["tools"] = chatConfig.Tools
if chatConfig.ToolChoice != nil {
reqBody["tool_choice"] = *chatConfig.ToolChoice
}
}
// extractToolCalls converts an OpenAI-compatible message's tool calls.
func extractToolCalls(message map[string]interface{}) []map[string]interface{} {
rawToolCalls, ok := message["tool_calls"].([]interface{})
if !ok {
return nil
}
toolCalls := make([]map[string]interface{}, 0, len(rawToolCalls))
for _, rawToolCall := range rawToolCalls {
if toolCall, ok := rawToolCall.(map[string]interface{}); ok {
toolCalls = append(toolCalls, toolCall)
}
}
return toolCalls
}
// setSortedToolCallsResult stores accumulated tool calls in index order.
func setSortedToolCallsResult(chatConfig *ChatConfig, accumulatedToolCalls map[int]map[string]any) {
if chatConfig == nil || len(accumulatedToolCalls) == 0 {
return
}
indices := make([]int, 0, len(accumulatedToolCalls))
for idx := range accumulatedToolCalls {
indices = append(indices, idx)
}
sort.Ints(indices)
toolCalls := make([]map[string]interface{}, 0, len(accumulatedToolCalls))
for _, idx := range indices {
toolCalls = append(toolCalls, accumulatedToolCalls[idx])
}
chatConfig.ToolCallsResult = &toolCalls
}
// accumulateToolCallDeltas merges streaming tool-call deltas by index.
func accumulateToolCallDeltas(delta map[string]interface{}, accumulatedToolCalls map[int]map[string]any) bool {
toolCallDeltas, ok := delta["tool_calls"].([]interface{})
if !ok {
return false
}
for _, toolCallDelta := range toolCallDeltas {
toolCall, ok := toolCallDelta.(map[string]interface{})
if !ok {
continue
}
idxF, ok := toolCall["index"].(float64)
if !ok {
continue
}
idx := int(idxF)
existing, hasExisting := accumulatedToolCalls[idx]
if !hasExisting {
accumulatedToolCalls[idx] = cloneMap(toolCall)
continue
}
appendStringField(existing, toolCall, "id")
if typ, ok := toolCall["type"].(string); ok && typ != "" {
existing["type"] = typ
}
mergeToolCallFunction(existing, toolCall)
}
return true
}
// appendStringField appends a non-empty string field from src into dst.
func appendStringField(dst, src map[string]interface{}, key string) {
value, ok := src[key].(string)
if !ok || value == "" {
return
}
if existing, ok := dst[key].(string); ok {
dst[key] = existing + value
} else {
dst[key] = value
}
}
// mergeToolCallFunction merges streamed function name and arguments.
func mergeToolCallFunction(existing, delta map[string]interface{}) {
fn, ok := delta["function"].(map[string]interface{})
if !ok {
return
}
existingFn, ok := existing["function"].(map[string]interface{})
if !ok {
existingFn = make(map[string]interface{})
existing["function"] = existingFn
}
appendStringField(existingFn, fn, "name")
appendStringField(existingFn, fn, "arguments")
}
// CloneMap returns a shallow copy of m.
func cloneMap(m map[string]interface{}) map[string]interface{} {
cp := make(map[string]interface{}, len(m))
for k, v := range m {
cp[k] = v
}
return cp
}