mirror of
https://github.com/infiniflow/ragflow.git
synced 2026-07-16 20:57:21 +08:00
06b07bbfd6
### What problem does this PR solve? Closes https://github.com/infiniflow/ragflow/issues/14571. Adds CAJAL as a first-class local scientific-writing option in RAGFlow: - registers `agnuxo/cajal-4b-p2pclaw` as a known Ollama chat model with a 32K context setting - adds a built-in “CAJAL scientific paper agent” template under the existing agent template catalog - preconfigures the agent for grounded scientific writing: retrieval first, citation traceability, LaTeX-ready output, and explicit limitations when evidence is missing - adds unit coverage to ensure the template normalizes through RAGFlow’s production template loader, keeps graph form data in sync, and exposes the Ollama model option Behavior/evidence gathered for the requested model: - Hugging Face model metadata for `Agnuxo/CAJAL-4B-P2PCLAW` reports `pipeline_tag=text-generation` and tags including `gguf`, `llama.cpp`, `vllm`, `scientific-research`, `papers`, `academic-writing`, `latex`, and `license:apache-2.0`. - The model card documents CAJAL as a 4B scientific paper generation model with 32K context, local inference, LaTeX/citation specialization, and CPU-only support around 5 tok/s on Ryzen 7 5800X. - Local CPU generation could not be completed on this machine because the advertised Ollama model name is not currently resolvable from Ollama’s registry: both `https://registry.ollama.ai/v2/agnuxo/cajal-4b-p2pclaw/manifests/latest` and `https://registry.ollama.ai/v2/library/agnuxo/cajal-4b-p2pclaw/manifests/latest` returned `404 Not Found`; the Hugging Face repo tree currently exposes an 8.4 GB `model.safetensors` but no GGUF artifact in `main`. The template therefore targets the documented Ollama model name for users who have the local CAJAL deployment/model file available. Verification run locally: ```bash python3 -m pytest test/test_cajal_template_unit.py -q # 3 passed in 0.34s python3 - <<'PY' import json, glob for f in sorted(glob.glob('agent/templates/*.json') + ['conf/llm_factories.json']): with open(f, encoding='utf-8') as fp: json.load(fp) print('json_ok') PY # json_ok python3 -m ruff check test/test_cajal_template_unit.py # All checks passed! git diff --check ``` `uv run pytest test/testcases/test_web_api/test_agent_app/test_cajal_template_unit.py -q` was also attempted first, but dependency setup failed before test collection while building `ormsgpack==1.5.0` from uv with a package metadata parse error. Clearing uv’s `ormsgpack` cache and retrying reproduced the same build failure, so the focused unit test was run with the system Python environment instead. ### Type of change - [ ] Bug Fix (non-breaking change which fixes an issue) - [x] New Feature (non-breaking change which adds functionality) - [ ] Documentation Update - [ ] Refactoring - [ ] Performance Improvement - [ ] Other (please describe): --------- Co-authored-by: sxxtony <sxxtony@users.noreply.github.com> Co-authored-by: yzc <yzc@users.noreply.github.com> Co-authored-by: Zhichang Yu <yuzhichang@gmail.com>
305 lines
16 KiB
JSON
305 lines
16 KiB
JSON
{
|
||
"id": "41",
|
||
"title": {
|
||
"en": "CAJAL scientific paper agent",
|
||
"de": "CAJAL-Agent für wissenschaftliche Arbeiten",
|
||
"zh": "CAJAL 科学论文助手"
|
||
},
|
||
"description": {
|
||
"en": "A local-first scientific paper generation agent for RAGFlow. It is preconfigured for Agnuxo/CAJAL-4B-P2PCLAW through Ollama, retrieves knowledge-base evidence, and drafts citation-grounded LaTeX-ready academic sections.",
|
||
"de": "Ein lokal ausgerichteter Agent zur Erstellung wissenschaftlicher Arbeiten in RAGFlow. Er ist für Agnuxo/CAJAL-4B-P2PCLAW über Ollama vorkonfiguriert, ruft Evidenz aus der Wissensdatenbank ab und erstellt zitationsgestützte, LaTeX-fähige akademische Abschnitte.",
|
||
"zh": "面向 RAGFlow 的本地优先科学论文生成助手。该模板预配置 Agnuxo/CAJAL-4B-P2PCLAW（Ollama），可检索知识库证据，并生成带引用依据、适合 LaTeX 的学术章节。"
|
||
},
|
||
"canvas_type": "Agent",
|
||
"canvas_types": [
|
||
"Agent",
|
||
"Recommended"
|
||
],
|
||
"dsl": {
|
||
"components": {
|
||
"Agent:NewPumasLick": {
|
||
"downstream": [
|
||
"Message:OrangeYearsShine"
|
||
],
|
||
"obj": {
|
||
"component_name": "Agent",
|
||
"params": {
|
||
"delay_after_error": 1,
|
||
"description": "",
|
||
"exception_comment": "",
|
||
"exception_default_value": "",
|
||
"exception_goto": [],
|
||
"exception_method": null,
|
||
"frequencyPenaltyEnabled": false,
|
||
"frequency_penalty": 0.5,
|
||
"llm_id": "agnuxo/cajal-4b-p2pclaw@Ollama",
|
||
"maxTokensEnabled": true,
|
||
"max_retries": 3,
|
||
"max_rounds": 3,
|
||
"max_tokens": 32768,
|
||
"mcp": [],
|
||
"message_history_window_size": 12,
|
||
"outputs": {
|
||
"content": {
|
||
"type": "string",
|
||
"value": ""
|
||
}
|
||
},
|
||
"parameter": "Precise",
|
||
"presencePenaltyEnabled": false,
|
||
"presence_penalty": 0.5,
|
||
"prompts": [
|
||
{
|
||
"role": "user",
|
||
"content": "# Research request\n{sys.query}\n\nUse the configured knowledge base retrieval tool before drafting. If no evidence is retrieved, state that limitation clearly."
|
||
}
|
||
],
|
||
"sys_prompt": "## Role & Task\nYou are **CAJAL**, a local-first scientific writing agent specialized in citation-grounded paper generation.\n\nUse RAGFlow retrieval results as the source of truth. Write precise academic content that can be pasted into a paper draft, technical report, or LaTeX manuscript. You are optimized for the local GGUF model `Agnuxo/CAJAL-4B-P2PCLAW` served through Ollama as `agnuxo/cajal-4b-p2pclaw`.\n\n## Operating Rules\n1. Decompose the user request into research goals, target section type, expected citation style, and missing evidence.\n2. Retrieve relevant knowledge-base passages before drafting factual claims.\n3. Ground every substantive claim in retrieved evidence. If evidence is missing, mark it as a limitation instead of inventing a citation.\n4. Prefer structured scientific writing: abstract, introduction, related work, methodology, experiments, results, limitations, and conclusion.\n5. Use LaTeX formatting for equations, symbols, algorithms, tables, and section headings when it helps the manuscript.\n6. Preserve traceability: cite source titles, document names, page numbers, or chunk identifiers when available in retrieved context.\n7. Keep language technical, concise, and reproducible. Avoid marketing language and vague generalizations.\n\n## Output Contract\nReturn one of the following, depending on the user request:\n- A complete paper section with citation markers and a short evidence map.\n- A literature review organized by themes, methods, findings, and gaps.\n- A methodology or experiment section with reproducible steps, assumptions, and limitations.\n- A LaTeX-ready abstract, introduction, or conclusion.\n\nAlways include:\n- **Draft**: the requested scientific content.\n- **Evidence used**: concise bullets mapping claims to retrieved sources.\n- **Limitations**: missing evidence, weak support, or assumptions that require verification.\n",
|
||
"temperature": 0.2,
|
||
"temperatureEnabled": true,
|
||
"tools": [
|
||
{
|
||
"component_name": "Retrieval",
|
||
"name": "Retrieval",
|
||
"params": {
|
||
"cross_languages": [],
|
||
"description": "Retrieve papers, reports, datasets, and notes that ground CAJAL scientific writing outputs.",
|
||
"empty_response": "",
|
||
"kb_ids": [],
|
||
"keywords_similarity_weight": 0.7,
|
||
"outputs": {
|
||
"formalized_content": {
|
||
"type": "string",
|
||
"value": ""
|
||
}
|
||
},
|
||
"rerank_id": "",
|
||
"similarity_threshold": 0.2,
|
||
"top_k": 1024,
|
||
"top_n": 10,
|
||
"use_kg": false
|
||
}
|
||
}
|
||
],
|
||
"topPEnabled": false,
|
||
"top_p": 0.75,
|
||
"user_prompt": "",
|
||
"visual_files_var": ""
|
||
}
|
||
},
|
||
"upstream": [
|
||
"begin"
|
||
]
|
||
},
|
||
"Message:OrangeYearsShine": {
|
||
"downstream": [],
|
||
"obj": {
|
||
"component_name": "Message",
|
||
"params": {
|
||
"content": [
|
||
"{Agent:NewPumasLick@content}"
|
||
]
|
||
}
|
||
},
|
||
"upstream": [
|
||
"Agent:NewPumasLick"
|
||
]
|
||
},
|
||
"begin": {
|
||
"downstream": [
|
||
"Agent:NewPumasLick"
|
||
],
|
||
"obj": {
|
||
"component_name": "Begin",
|
||
"params": {
|
||
"enablePrologue": true,
|
||
"inputs": {},
|
||
"mode": "conversational",
|
||
"prologue": "Hi, I am CAJAL in RAGFlow. Add research papers or datasets to your knowledge base, then ask me to draft a citation-grounded paper section, literature review, methodology, or LaTeX-ready abstract."
|
||
}
|
||
},
|
||
"upstream": []
|
||
}
|
||
},
|
||
"globals": {
|
||
"sys.conversation_turns": 0,
|
||
"sys.files": [],
|
||
"sys.query": "",
|
||
"sys.user_id": ""
|
||
},
|
||
"graph": {
|
||
"edges": [
|
||
{
|
||
"data": {
|
||
"isHovered": false
|
||
},
|
||
"id": "xy-edge__beginstart-Agent:NewPumasLickend",
|
||
"source": "begin",
|
||
"sourceHandle": "start",
|
||
"target": "Agent:NewPumasLick",
|
||
"targetHandle": "end"
|
||
},
|
||
{
|
||
"data": {
|
||
"isHovered": false
|
||
},
|
||
"id": "xy-edge__Agent:NewPumasLickstart-Message:OrangeYearsShineend",
|
||
"markerEnd": "logo",
|
||
"source": "Agent:NewPumasLick",
|
||
"sourceHandle": "start",
|
||
"style": {
|
||
"stroke": "rgba(91, 93, 106, 1)",
|
||
"strokeWidth": 1
|
||
},
|
||
"target": "Message:OrangeYearsShine",
|
||
"targetHandle": "end",
|
||
"type": "buttonEdge",
|
||
"zIndex": 1001
|
||
}
|
||
],
|
||
"nodes": [
|
||
{
|
||
"data": {
|
||
"form": {
|
||
"enablePrologue": true,
|
||
"inputs": {},
|
||
"mode": "conversational",
|
||
"prologue": "Hi, I am CAJAL in RAGFlow. Add research papers or datasets to your knowledge base, then ask me to draft a citation-grounded paper section, literature review, methodology, or LaTeX-ready abstract."
|
||
},
|
||
"label": "Begin",
|
||
"name": "begin"
|
||
},
|
||
"dragging": false,
|
||
"id": "begin",
|
||
"measured": {
|
||
"height": 48,
|
||
"width": 200
|
||
},
|
||
"position": {
|
||
"x": -9.569875358221438,
|
||
"y": 205.84018385864917
|
||
},
|
||
"selected": false,
|
||
"sourcePosition": "left",
|
||
"targetPosition": "right",
|
||
"type": "beginNode"
|
||
},
|
||
{
|
||
"data": {
|
||
"form": {
|
||
"content": [
|
||
"{Agent:NewPumasLick@content}"
|
||
]
|
||
},
|
||
"label": "Scientific Draft",
|
||
"name": "Response"
|
||
},
|
||
"dragging": false,
|
||
"id": "Message:OrangeYearsShine",
|
||
"measured": {
|
||
"height": 56,
|
||
"width": 200
|
||
},
|
||
"position": {
|
||
"x": 734.4061285881053,
|
||
"y": 199.9706031723009
|
||
},
|
||
"selected": false,
|
||
"sourcePosition": "right",
|
||
"targetPosition": "left",
|
||
"type": "messageNode"
|
||
},
|
||
{
|
||
"data": {
|
||
"form": {
|
||
"delay_after_error": 1,
|
||
"description": "",
|
||
"exception_comment": "",
|
||
"exception_default_value": "",
|
||
"exception_goto": [],
|
||
"exception_method": null,
|
||
"frequencyPenaltyEnabled": false,
|
||
"frequency_penalty": 0.5,
|
||
"llm_id": "agnuxo/cajal-4b-p2pclaw@Ollama",
|
||
"maxTokensEnabled": true,
|
||
"max_retries": 3,
|
||
"max_rounds": 3,
|
||
"max_tokens": 32768,
|
||
"mcp": [],
|
||
"message_history_window_size": 12,
|
||
"outputs": {
|
||
"content": {
|
||
"type": "string",
|
||
"value": ""
|
||
}
|
||
},
|
||
"parameter": "Precise",
|
||
"presencePenaltyEnabled": false,
|
||
"presence_penalty": 0.5,
|
||
"prompts": [
|
||
{
|
||
"role": "user",
|
||
"content": "# Research request\n{sys.query}\n\nUse the configured knowledge base retrieval tool before drafting. If no evidence is retrieved, state that limitation clearly."
|
||
}
|
||
],
|
||
"sys_prompt": "## Role & Task\nYou are **CAJAL**, a local-first scientific writing agent specialized in citation-grounded paper generation.\n\nUse RAGFlow retrieval results as the source of truth. Write precise academic content that can be pasted into a paper draft, technical report, or LaTeX manuscript. You are optimized for the local GGUF model `Agnuxo/CAJAL-4B-P2PCLAW` served through Ollama as `agnuxo/cajal-4b-p2pclaw`.\n\n## Operating Rules\n1. Decompose the user request into research goals, target section type, expected citation style, and missing evidence.\n2. Retrieve relevant knowledge-base passages before drafting factual claims.\n3. Ground every substantive claim in retrieved evidence. If evidence is missing, mark it as a limitation instead of inventing a citation.\n4. Prefer structured scientific writing: abstract, introduction, related work, methodology, experiments, results, limitations, and conclusion.\n5. Use LaTeX formatting for equations, symbols, algorithms, tables, and section headings when it helps the manuscript.\n6. Preserve traceability: cite source titles, document names, page numbers, or chunk identifiers when available in retrieved context.\n7. Keep language technical, concise, and reproducible. Avoid marketing language and vague generalizations.\n\n## Output Contract\nReturn one of the following, depending on the user request:\n- A complete paper section with citation markers and a short evidence map.\n- A literature review organized by themes, methods, findings, and gaps.\n- A methodology or experiment section with reproducible steps, assumptions, and limitations.\n- A LaTeX-ready abstract, introduction, or conclusion.\n\nAlways include:\n- **Draft**: the requested scientific content.\n- **Evidence used**: concise bullets mapping claims to retrieved sources.\n- **Limitations**: missing evidence, weak support, or assumptions that require verification.\n",
|
||
"temperature": 0.2,
|
||
"temperatureEnabled": true,
|
||
"tools": [
|
||
{
|
||
"component_name": "Retrieval",
|
||
"name": "Retrieval",
|
||
"params": {
|
||
"cross_languages": [],
|
||
"description": "Retrieve papers, reports, datasets, and notes that ground CAJAL scientific writing outputs.",
|
||
"empty_response": "",
|
||
"kb_ids": [],
|
||
"keywords_similarity_weight": 0.7,
|
||
"outputs": {
|
||
"formalized_content": {
|
||
"type": "string",
|
||
"value": ""
|
||
}
|
||
},
|
||
"rerank_id": "",
|
||
"similarity_threshold": 0.2,
|
||
"top_k": 1024,
|
||
"top_n": 10,
|
||
"use_kg": false
|
||
}
|
||
}
|
||
],
|
||
"topPEnabled": false,
|
||
"top_p": 0.75,
|
||
"user_prompt": "",
|
||
"visual_files_var": ""
|
||
},
|
||
"label": "CAJAL Writer",
|
||
"name": "Knowledge Base Agent"
|
||
},
|
||
"dragging": false,
|
||
"id": "Agent:NewPumasLick",
|
||
"measured": {
|
||
"height": 84,
|
||
"width": 200
|
||
},
|
||
"position": {
|
||
"x": 347.00048227952215,
|
||
"y": 186.49109364794631
|
||
},
|
||
"selected": false,
|
||
"sourcePosition": "right",
|
||
"targetPosition": "left",
|
||
"type": "agentNode"
|
||
}
|
||
]
|
||
},
|
||
"history": [],
|
||
"memory": [],
|
||
"messages": [],
|
||
"path": [],
|
||
"retrieval": []
|
||
},
|
||
"avatar": "data:image/png;base64,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"
|
||
}
|