mirror of
https://github.com/infiniflow/ragflow.git
synced 2026-07-03 01:01:56 +08:00
3fa15c0e2f
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
454 lines
16 KiB
JSON
454 lines
16 KiB
JSON
{
|
|
"components": {
|
|
"Begin:7d83abf3b4f611efa3c40242ac120002": {
|
|
"downstream": ["Categorize:11111111aaaa0001aaaa0001aaaa0001"],
|
|
"upstream": [],
|
|
"obj": {
|
|
"component_name": "Begin",
|
|
"params": {
|
|
"mode": "conversational",
|
|
"enablePrologue": true,
|
|
"prologue": "Hi! I'm your multi-source research copilot. Ask me anything and I'll dig through our knowledge base, score the evidence, and synthesize a sourced answer.",
|
|
"inputs": [
|
|
{"name": "query", "type": "string", "value": "{sys.query}"},
|
|
{"name": "user_id", "type": "string", "value": "{sys.user_id}"},
|
|
{"name": "session_id", "type": "string", "value": "{sys.session_id}"}
|
|
],
|
|
"outputs": []
|
|
}
|
|
}
|
|
},
|
|
"Categorize:11111111aaaa0001aaaa0001aaaa0001": {
|
|
"downstream": ["Switch:22222222bbbb0001bbbb0001bbbb0001"],
|
|
"obj": {
|
|
"component_name": "Categorize",
|
|
"params": {
|
|
"category_description": {
|
|
"research": {
|
|
"to": "Switch:22222222bbbb0001bbbb0001bbbb0001",
|
|
"description": "User is asking a research question, a how-to, a comparison, or wants a sourced answer.",
|
|
"examples": [
|
|
"What is retrieval-augmented generation?",
|
|
"Compare LoRA vs full fine-tuning",
|
|
"Summarize the latest papers on agent loops"
|
|
]
|
|
},
|
|
"clarify": {
|
|
"to": "UserFillup:cccccccc00010001cccc00010001cccc0001",
|
|
"description": "User's intent is ambiguous and we need them to disambiguate.",
|
|
"examples": ["explain that", "tell me about it", "go on"]
|
|
},
|
|
"decline": {
|
|
"to": "Message:00000000eeee0001eeee0001eeee0001",
|
|
"description": "User is asking for content we cannot provide (e.g. medical diagnosis, legal advice, or unrelated chitchat).",
|
|
"examples": ["diagnose my symptoms", "should I sue my employer", "hi how are you"]
|
|
}
|
|
},
|
|
"llm_id": "gpt-4o-mini",
|
|
"query": "{Begin:7d83abf3b4f611efa3c40242ac120002@query}",
|
|
"message_history_window_size": 5,
|
|
"temperature": 0.0,
|
|
"outputs": [{"name": "intent", "type": "string"}]
|
|
}
|
|
}
|
|
},
|
|
"Switch:22222222bbbb0001bbbb0001bbbb0001": {
|
|
"downstream": [
|
|
"Agent:33333333cccc0001cccc0001cccc0001",
|
|
"UserFillup:cccccccc00010001cccc00010001cccc0001",
|
|
"Message:00000000eeee0001eeee0001eeee0001"
|
|
],
|
|
"obj": {
|
|
"component_name": "Switch",
|
|
"params": {
|
|
"logical_operator": "or",
|
|
"conditions": [
|
|
{
|
|
"cpn_id": "Categorize:11111111aaaa0001aaaa0001aaaa0001",
|
|
"operator": "=",
|
|
"value": "research",
|
|
"to": "Agent:33333333cccc0001cccc0001cccc0001"
|
|
}
|
|
],
|
|
"end_cpn_ids": [
|
|
"UserFillup:cccccccc00010001cccc00010001cccc0001",
|
|
"Message:00000000eeee0001eeee0001eeee0001"
|
|
],
|
|
"operators": [
|
|
"contains", "not contains", "start with", "end with",
|
|
"empty", "not empty", "=", "≠", ">", "<", "≥", "≤"
|
|
]
|
|
}
|
|
}
|
|
},
|
|
"UserFillup:cccccccc00010001cccc00010001cccc0001": {
|
|
"downstream": ["Message:00000000eeee0002eeee0002eeee0002"],
|
|
"obj": {
|
|
"component_name": "UserFillup",
|
|
"params": {
|
|
"inputs": [
|
|
{
|
|
"name": "intent",
|
|
"type": "select",
|
|
"options": ["research", "chitchat", "support"],
|
|
"value": "research"
|
|
},
|
|
{
|
|
"name": "clarifying_question",
|
|
"type": "string",
|
|
"value": "Could you tell me a bit more about what you want to know? I can answer research questions, summarize documents, or compare approaches."
|
|
}
|
|
],
|
|
"outputs": [
|
|
{"name": "intent", "type": "string"},
|
|
{"name": "clarifying_question", "type": "string"}
|
|
]
|
|
}
|
|
}
|
|
},
|
|
"Message:00000000eeee0001eeee0001eeee0001": {
|
|
"downstream": [],
|
|
"obj": {
|
|
"component_name": "Message",
|
|
"params": {
|
|
"content": "I'm sorry, but I can't help with that kind of request. I can answer research questions, summarize documents, or compare approaches — feel free to ask one of those!",
|
|
"outputs": []
|
|
}
|
|
}
|
|
},
|
|
"Message:00000000eeee0002eeee0002eeee0002": {
|
|
"downstream": [],
|
|
"obj": {
|
|
"component_name": "Message",
|
|
"params": {
|
|
"content": "Got it. To help me route your question, please share a clarifying question via the form.",
|
|
"outputs": []
|
|
}
|
|
}
|
|
},
|
|
"Agent:33333333cccc0001cccc0001cccc0001": {
|
|
"downstream": ["Loop:dddddddd0001000100010001000100010001"],
|
|
"obj": {
|
|
"component_name": "Agent",
|
|
"params": {
|
|
"description": "Decomposes the user's research question into 1-5 self-contained sub-questions that the loop can answer one at a time.",
|
|
"llm_id": "gpt-4o",
|
|
"prompts": [
|
|
{
|
|
"role": "system",
|
|
"content": "You are a meticulous research planner. Decompose the user's question into 1-5 self-contained sub-questions that can each be answered by retrieving from a knowledge base. Output JSON: {\"sub_questions\": [\"...\", \"...\"]}."
|
|
},
|
|
{
|
|
"role": "user",
|
|
"content": "{Begin:7d83abf3b4f611efa3c40242ac120002@query}"
|
|
}
|
|
],
|
|
"tools": ["retrieval"],
|
|
"memory": [
|
|
{"role": "user", "limit": 6, "session_only": true}
|
|
],
|
|
"cite": false,
|
|
"max_steps": 4,
|
|
"delay_after_error": 2,
|
|
"exception_method": "default",
|
|
"exception_default_value": "{\"sub_questions\": [\"{Begin:7d83abf3b4f611efa3c40242ac120002@query}\"]}",
|
|
"exception_goto": "Loop:dddddddd0001000100010001000100010001",
|
|
"exception_comment": "fallback to a single-sub-question loop with the original query",
|
|
"outputs": [
|
|
{"name": "sub_questions", "type": "json"}
|
|
]
|
|
}
|
|
}
|
|
},
|
|
"Loop:dddddddd0001000100010001000100010001": {
|
|
"downstream": ["Iteration:7777777700010001000100010001aaaa0001"],
|
|
"obj": {
|
|
"component_name": "Loop",
|
|
"params": {
|
|
"loop_variables": [
|
|
{"name": "accumulated_evidence", "source": "literal", "value": "[]"}
|
|
],
|
|
"loop_termination_condition": [
|
|
{
|
|
"variable": "accumulated_evidence",
|
|
"operator": "≥",
|
|
"value": "5"
|
|
}
|
|
],
|
|
"maximum_loop_count": 5,
|
|
"outputs": []
|
|
}
|
|
}
|
|
},
|
|
"LoopItem:eeeeeeee00010001000100010001bbbb0001": {
|
|
"downstream": ["Retrieval:555555550001000100010001eeee0001"],
|
|
"obj": {
|
|
"component_name": "LoopItem",
|
|
"params": {
|
|
"outputs": []
|
|
}
|
|
}
|
|
},
|
|
"Retrieval:555555550001000100010001eeee0001": {
|
|
"downstream": ["LLM:444444440001000100010001dddd0001"],
|
|
"obj": {
|
|
"component_name": "Retrieval",
|
|
"params": {
|
|
"kb_ids": ["kb_research_2024", "kb_company_wiki", "kb_industry_reports"],
|
|
"top_k": 10,
|
|
"rerank_id": "bge-reranker-v2-m3",
|
|
"empty_response": "No matching documents were found.",
|
|
"keywords": [
|
|
"{Agent:33333333cccc0001cccc0001cccc0001@sub_questions}"
|
|
],
|
|
"query": "{Begin:7d83abf3b4f611efa3c40242ac120002@query}",
|
|
"outputs": [{"name": "chunks", "type": "json"}]
|
|
}
|
|
}
|
|
},
|
|
"LLM:444444440001000100010001dddd0001": {
|
|
"downstream": ["Categorize:11111111aaaa0002aaaa0002aaaa0002"],
|
|
"obj": {
|
|
"component_name": "LLM",
|
|
"params": {
|
|
"model": "gpt-4o-mini",
|
|
"temperature": 0.0,
|
|
"system": "You are a relevance scorer. For each chunk, output a JSON object {\"idx\": <int>, \"score\": <0.0-1.0>, \"reason\": \"<short>\"}.",
|
|
"prompt": "Score the relevance of each chunk to the user's question.\n\nQuestion: {Begin:7d83abf3b4f611efa3c40242ac120002@query}\n\nChunks: {Retrieval:555555550001000100010001eeee0001@chunks}",
|
|
"max_tokens": 1024,
|
|
"outputs": [{"name": "scored_chunks", "type": "json"}]
|
|
}
|
|
}
|
|
},
|
|
"Categorize:11111111aaaa0002aaaa0002aaaa0002": {
|
|
"downstream": ["Switch:22222222bbbb0002bbbb0002bbbb0002"],
|
|
"obj": {
|
|
"component_name": "Categorize",
|
|
"params": {
|
|
"category_description": {
|
|
"high_relevance": {
|
|
"to": "Switch:22222222bbbb0002bbbb0002bbbb0002",
|
|
"description": "The retrieved chunk is highly relevant to the user's question (score ≥ 0.7).",
|
|
"examples": ["directly answers the question", "contains the exact term the user asked about"]
|
|
},
|
|
"low_relevance": {
|
|
"to": "VariableAggregator:666666660001000100010001ffff0001",
|
|
"description": "The retrieved chunk is off-topic or redundant (score < 0.3).",
|
|
"examples": ["tangentially mentions the topic", "duplicates another chunk"]
|
|
}
|
|
},
|
|
"llm_id": "gpt-4o-mini",
|
|
"query": "{LLM:444444440001000100010001dddd0001@scored_chunks}",
|
|
"message_history_window_size": 1,
|
|
"temperature": 0.0,
|
|
"outputs": [{"name": "relevance", "type": "string"}]
|
|
}
|
|
}
|
|
},
|
|
"Switch:22222222bbbb0002bbbb0002bbbb0002": {
|
|
"downstream": ["VariableAggregator:666666660001000100010001ffff0001"],
|
|
"obj": {
|
|
"component_name": "Switch",
|
|
"params": {
|
|
"logical_operator": "or",
|
|
"conditions": [
|
|
{
|
|
"cpn_id": "Categorize:11111111aaaa0002aaaa0002aaaa0002",
|
|
"operator": "contains",
|
|
"value": "high_relevance",
|
|
"to": "VariableAggregator:666666660001000100010001ffff0001"
|
|
}
|
|
],
|
|
"end_cpn_ids": ["VariableAggregator:666666660001000100010001ffff0001"],
|
|
"operators": [
|
|
"contains", "not contains", "=", "≠"
|
|
]
|
|
}
|
|
}
|
|
},
|
|
"VariableAggregator:666666660001000100010001ffff0001": {
|
|
"downstream": ["Iteration:7777777700010001000100010001aaaa0001"],
|
|
"obj": {
|
|
"component_name": "VariableAggregator",
|
|
"params": {
|
|
"groups": [
|
|
{
|
|
"group_id": "evidence_group",
|
|
"group_name": "evidence",
|
|
"llm_id": "gpt-4o-mini",
|
|
"format": "{Retrieval:555555550001000100010001eeee0001@chunks}",
|
|
"refusal_answer": "no evidence",
|
|
"output": {
|
|
"value": {"type": "collected", "value": "evidence"}
|
|
}
|
|
}
|
|
],
|
|
"outputs": [
|
|
{"name": "evidence", "type": "json"}
|
|
]
|
|
}
|
|
}
|
|
},
|
|
"Iteration:7777777700010001000100010001aaaa0001": {
|
|
"downstream": ["LLM:444444440002000200020002dddd0002"],
|
|
"obj": {
|
|
"component_name": "Iteration",
|
|
"params": {
|
|
"items_ref": "VariableAggregator:666666660001000100010001ffff0001@evidence",
|
|
"variable": {
|
|
"chunk_id": "",
|
|
"chunk_text": ""
|
|
},
|
|
"outputs": []
|
|
}
|
|
}
|
|
},
|
|
"IterationItem:8888888800010001000100010001bbbb0001": {
|
|
"downstream": ["VariableAggregator:666666660002000200020002ffff0002"],
|
|
"obj": {
|
|
"component_name": "IterationItem",
|
|
"params": {
|
|
"outputs": [
|
|
{"name": "excerpt", "type": "string"}
|
|
]
|
|
}
|
|
}
|
|
},
|
|
"VariableAggregator:666666660002000200020002ffff0002": {
|
|
"downstream": ["LLM:444444440002000200020002dddd0002"],
|
|
"obj": {
|
|
"component_name": "VariableAggregator",
|
|
"params": {
|
|
"groups": [
|
|
{
|
|
"group_id": "ranked_group",
|
|
"group_name": "ranked_excerpts",
|
|
"llm_id": "gpt-4o-mini",
|
|
"format": "{IterationItem:8888888800010001000100010001bbbb0001@excerpt}",
|
|
"refusal_answer": "no excerpt",
|
|
"output": {
|
|
"value": {"type": "collected", "value": "ranked_excerpts"}
|
|
}
|
|
}
|
|
],
|
|
"outputs": [
|
|
{"name": "ranked_excerpts", "type": "json"}
|
|
]
|
|
}
|
|
}
|
|
},
|
|
"LLM:444444440002000200020002dddd0002": {
|
|
"downstream": ["CodeExec:bbbbbbbb0001000100010001cccc0001"],
|
|
"obj": {
|
|
"component_name": "LLM",
|
|
"params": {
|
|
"model": "deepseek-chat",
|
|
"temperature": 0.2,
|
|
"system": "You are a careful synthesizer. Combine the ranked excerpts into a sourced answer. Cite the chunk id for every claim.",
|
|
"prompt": "Question: {Begin:7d83abf3b4f611efa3c40242ac120002@query}\n\nRanked excerpts: {VariableAggregator:666666660002000200020002ffff0002@ranked_excerpts}",
|
|
"max_tokens": 1024,
|
|
"outputs": [
|
|
{"name": "draft_answer", "type": "string"},
|
|
{"name": "citations", "type": "json"}
|
|
]
|
|
}
|
|
}
|
|
},
|
|
"CodeExec:bbbbbbbb0001000100010001cccc0001": {
|
|
"downstream": ["Switch:22222222bbbb0003bbbb0003bbbb0003"],
|
|
"obj": {
|
|
"component_name": "CodeExec",
|
|
"params": {
|
|
"script": "import json, sys\ndraft = sys.argv[1]\ncitations = json.loads(sys.argv[2])\n# Validate: every claim in the draft must reference a real citation id.\nok = all(c.get('id') for c in citations) and len(citations) > 0\nprint(json.dumps({\"ok\": ok, \"citation_count\": len(citations), \"draft_len\": len(draft)}))",
|
|
"lang": "python",
|
|
"arguments": [
|
|
"{LLM:444444440002000200020002dddd0002@draft_answer}",
|
|
"{LLM:444444440002000200020002dddd0002@citations}"
|
|
],
|
|
"outputs": [
|
|
{"name": "validation", "type": "json"}
|
|
]
|
|
}
|
|
}
|
|
},
|
|
"Switch:22222222bbbb0003bbbb0003bbbb0003": {
|
|
"downstream": [
|
|
"VariableAssigner:9999999900010001000100010001cccc0001",
|
|
"LLM:444444440002000200020002dddd0002",
|
|
"Message:00000000eeee0003eeee0003eeee0003"
|
|
],
|
|
"obj": {
|
|
"component_name": "Switch",
|
|
"params": {
|
|
"logical_operator": "or",
|
|
"conditions": [
|
|
{
|
|
"cpn_id": "CodeExec:bbbbbbbb0001000100010001cccc0001",
|
|
"operator": "contains",
|
|
"value": "\"ok\": true",
|
|
"to": "VariableAssigner:9999999900010001000100010001cccc0001"
|
|
}
|
|
],
|
|
"end_cpn_ids": [
|
|
"LLM:444444440002000200020002dddd0002",
|
|
"Message:00000000eeee0003eeee0003eeee0003"
|
|
],
|
|
"operators": [
|
|
"contains", "not contains", "=", "≠"
|
|
]
|
|
}
|
|
}
|
|
},
|
|
"VariableAssigner:9999999900010001000100010001cccc0001": {
|
|
"downstream": ["Message:00000000eeee0004eeee0004eeee0004"],
|
|
"obj": {
|
|
"component_name": "VariableAssigner",
|
|
"params": {
|
|
"concurrent": true,
|
|
"orders": [
|
|
{
|
|
"id": "research_done",
|
|
"name": "research_done",
|
|
"source": "literal",
|
|
"value": "true"
|
|
},
|
|
{
|
|
"id": "draft_answer",
|
|
"name": "draft_answer",
|
|
"source": "ref",
|
|
"ref": "{LLM:444444440002000200020002dddd0002@draft_answer}"
|
|
},
|
|
{
|
|
"id": "citations",
|
|
"name": "citations",
|
|
"source": "ref",
|
|
"ref": "{LLM:444444440002000200020002dddd0002@citations}"
|
|
}
|
|
],
|
|
"outputs": []
|
|
}
|
|
}
|
|
},
|
|
"Message:00000000eeee0003eeee0003eeee0003": {
|
|
"downstream": [],
|
|
"obj": {
|
|
"component_name": "Message",
|
|
"params": {
|
|
"content": "I couldn't validate the answer after multiple attempts. Here is my best-effort draft:\n\n{LLM:444444440002000200020002dddd0002@draft_answer}",
|
|
"outputs": []
|
|
}
|
|
}
|
|
},
|
|
"Message:00000000eeee0004eeee0004eeee0004": {
|
|
"downstream": [],
|
|
"obj": {
|
|
"component_name": "Message",
|
|
"params": {
|
|
"content": "{LLM:444444440002000200020002dddd0002@draft_answer}\n\nCitations:\n{LLM:444444440002000200020002dddd0002@citations}",
|
|
"outputs": []
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|