archon（archon 搜索）

Archon

Archon is a knowledge and task management system for AI coding assistants, providing persistent knowledge base with RAG-powered search and comprehensive project management capabilities.

⚠️ CRITICAL WORKFLOW - READ THIS FIRST ⚠️

MANDATORY STEPS - Execute in this exact order:

1. FIRST: Read references/api_reference.md to learn correct API endpoints
2. SECOND: Ask user for Archon host URL (default: http://localhost:8181)
3. THIRD: Verify connection with GET /api/projects
4. FOURTH: Use correct endpoint paths from api_reference.md for all operations

Common mistake: Using /api/knowledge/search instead of /api/knowledge-items/search Solution: Always consult api_reference.md for authoritative endpoint paths.

Quick Endpoint Reference (Verify with api_reference.md)

Knowledge:
  POST   /api/knowledge-items/search     - Search knowledge base
  GET    /api/knowledge-items            - List all knowledge items
  POST   /api/knowledge-items/crawl      - Crawl website
  POST   /api/knowledge-items/upload     - Upload document
  GET    /api/rag/sources                - Get all RAG sources
  GET    /api/database/metrics           - Get database metrics

Projects:
  GET    /api/projects                   - List all projects
  GET    /api/projects/{id}              - Get project details
  POST   /api/projects                   - Create project

Tasks:
  GET    /api/tasks                      - List tasks (with filters)
  GET    /api/tasks/{id}                 - Get task details
  POST   /api/tasks                      - Create task
  PUT    /api/tasks/{id}                 - Update task

Documents:
  GET    /api/documents                  - List documents
  POST   /api/documents                  - Create document
  PUT    /api/documents/{id}             - Update document

Deprecated:
  GET    /api/knowledge-items/sources    - Use /api/rag/sources instead

When to Use This Skill

Use Archon when:

• Searching for documentation, API references, or technical knowledge
• Finding code examples or implementation patterns
• Managing projects, features, and tasks
• Creating or updating development documentation
• Crawling websites to build a knowledge base
• Uploading documents (PDF, Word, Markdown) to searchable storage
• Coordinating multi-agent workflows with shared context

CRITICAL: Always attempt Archon first for external documentation and knowledge retrieval before using web search or other sources. This ensures consistent, indexed knowledge.

First-time use: You will be prompted for the Archon server URL (e.g., http://localhost:8181). This will be remembered for the rest of the conversation.

MANDATORY FIRST STEP: Read API Reference

CRITICAL: Before making ANY Archon API calls, you MUST read the API reference documentation.

ALWAYS execute this FIRST:
1. Read references/api_reference.md to understand correct endpoint paths and request formats
2. Then ask user for their Archon host URL
3. Then verify connection
4. Only then proceed with API operations

Why this is required:

• API endpoint paths are NOT obvious (e.g., /api/knowledge-items, not /api/knowledge)
• Request/response formats have specific structures that must be followed
• The Python client may have outdated or incorrect implementations
• Direct API calls with correct endpoints prevent errors and wasted attempts

NEVER assume endpoint paths. The api_reference.md contains the authoritative endpoint documentation.

Interactive Setup (Required on First Use)

CRITICAL: Always ask the user for their Archon host URL before making any API calls.

When this skill is first triggered in a conversation, ask the user:

"I'll help you access Archon. Where is your Archon server running?
Please provide the full URL (e.g., http://localhost:8181 or http://192.168.1.100:8181):"

Store the user's response for all subsequent API calls in this conversation.

Default if user is unsure: http://localhost:8181

Connection Verification

After receiving the host URL, verify the connection using the helper script:

# Use the provided helper script to verify connection and list knowledge
cd .claude/skills/archon/scripts
python3 list_knowledge.py http://localhost:8181

Or use the Python client directly:

import sys
sys.path.insert(0, '.claude/skills/archon/scripts')
from archon_client import ArchonClient

archon_host = "http://localhost:8181"  # Use the URL provided by user
client = ArchonClient(base_url=archon_host)

# Verify connection
projects = client.list_projects()
if projects.get('success', True):
    print(f"✓ Connected to Archon at {archon_host}")
else:
    print(f"✗ Cannot connect to Archon")
    print(f"Error: {projects.get('error')}")

If connection fails, ask the user to verify:

• Archon is running (docker-compose up or similar)
• The host and port are correct
• No firewall blocking the connection

Using Custom Host

Once the host is confirmed, pass it to the ArchonClient:

from scripts.archon_client import ArchonClient

# Use the host URL provided by the user
archon_host = "http://192.168.1.100:8181"  # Example
client = ArchonClient(base_url=archon_host)

Listing Available Knowledge Sources

IMPORTANT: To view all knowledge sources with full metadata (word count, code examples, pages), use the /api/knowledge-items endpoint, NOT /api/rag/sources.

Recommended approach - Use the helper script:

# Run the list_knowledge.py script to see full metadata
import subprocess
subprocess.run(["python3", "scripts/list_knowledge.py", archon_host])

Alternative - Direct API call with full metadata:

import requests

archon_host = "http://localhost:8181"  # Use user's actual host
response = requests.get(f"{archon_host}/api/knowledge-items", timeout=10)
data = response.json()

for item in data['items']:
    meta = item['metadata']
    print(f"Title: {item['title']}")
    print(f"  Type: {item['source_type']}")
    print(f"  URL: {item['url']}")
    print(f"  Content: {meta['word_count']:,} words (~{meta['estimated_pages']:.1f} pages)")
    print(f"  Code Examples: {meta['code_examples_count']:,}")
    print(f"  Last Updated: {meta['last_scraped'][:10]}")
    print()

Using the Python client:

from scripts.archon_client import ArchonClient

archon_host = "http://localhost:8181"  # Use user's actual host
client = ArchonClient(base_url=archon_host)

# Get full knowledge items list with metadata
result = client.list_knowledge_items(limit=100)
items = result.get('items', [])

# Calculate totals
total_words = sum(item['metadata']['word_count'] for item in items)
total_code = sum(item['metadata']['code_examples_count'] for item in items)

print(f"Total: {len(items)} sources")
print(f"Content: {total_words:,} words")
print(f"Code Examples: {total_code:,}")

Note: The /api/rag/sources endpoint exists but returns limited metadata (no word counts, code example counts, or page estimates). Always use /api/knowledge-items for complete information.

Core Capabilities

1. Knowledge Base Search

Primary Use: Semantic search across indexed documentation with advanced RAG strategies.

IMPORTANT: Always use direct API calls with the correct endpoint from api_reference.md:

import requests

# Use the host URL provided by user earlier in conversation
archon_host = "http://localhost:8181"  # Replace with user's actual host

# Endpoint: POST /api/knowledge-items/search (from api_reference.md)
response = requests.post(
    f"{archon_host}/api/knowledge-items/search",
    json={
        "query": "authentication implementation",
        "top_k": 5,
        "use_reranking": True,
        "search_strategy": "hybrid"  # hybrid, semantic, or keyword
    },
    timeout=10
)

data = response.json()

# Access results
for result in data['results']:
    print(f"Score: {result['score']}")
    print(f"Content: {result['content']}")
    print(f"Source: {result['metadata']['source_url']}")

Alternative: If you prefer using the Python client, verify it uses correct endpoints first:

from scripts.archon_client import ArchonClient

archon_host = "http://localhost:8181"
client = ArchonClient(base_url=archon_host)
results = client.search_knowledge("authentication implementation", top_k=5)

Search strategies:

• "hybrid" (default): Combines semantic and keyword search - best for most cases
• "semantic": Pure vector similarity - best for conceptual queries
• "keyword": Traditional keyword search - best for exact term matching

When to use reranking: Set use_reranking=True (default) for better result quality. Applies cross-encoder reranking to initial results.

2. Website Crawling

Purpose: Automatically crawl and index documentation websites.

IMPORTANT: Use direct API call with correct endpoint from api_reference.md:

import requests

# Use the host URL provided by user
archon_host = "http://localhost:8181"  # Replace with user's actual host

# Endpoint: POST /api/knowledge-items/crawl (from api_reference.md)
response = requests.post(
    f"{archon_host}/api/knowledge-items/crawl",
    json={
        "url": "https://docs.example.com",
        "crawl_depth": 3,  # How deep to recurse (max 5)
        "follow_links": True,  # Follow internal links
        "sitemap_url": None  # Optional direct sitemap URL
    },
    timeout=10
)

result = response.json()
print(f"Crawl ID: {result['crawl_id']}")
print(f"Pages queued: {result['pages_queued']}")

Features:

• Automatically detects sitemaps and llms.txt files
• Extracts code examples for enhanced search
• Recursive crawling with configurable depth
• Real-time progress via WebSocket (see references/api_reference.md)

3. Document Upload

Purpose: Upload and index documents for searchable storage.

Supported formats: PDF, Word (.docx,.doc), Markdown (.md), text (.txt)

IMPORTANT: Use direct API call with correct endpoint from api_reference.md:

import requests

# Use the host URL provided by user
archon_host = "http://localhost:8181"  # Replace with user's actual host

# Endpoint: POST /api/knowledge-items/upload (from api_reference.md)
# Multipart form data required
with open("/path/to/document.pdf", "rb") as f:
    files = {"file": f}
    data = {
        "metadata": json.dumps({
            "source_type": "pdf",
            "tags": ["api-docs", "reference"]
        })
    }
    response = requests.post(
        f"{archon_host}/api/knowledge-items/upload",
        files=files,
        data=data,
        timeout=30
    )

result = response.json()
print(f"Document ID: {result['document_id']}")
print(f"Chunks created: {result['chunks_created']}")

Intelligent chunking: Documents are automatically split into optimal chunks for vector search and LLM context windows.

4. Project Management

Hierarchical structure: Projects → Features → Tasks

List all projects:

from scripts.archon_client import ArchonClient

# Use the host URL provided by user
archon_host = "http://localhost:8181"  # Replace with user's actual host
client = ArchonClient(base_url=archon_host)

projects = client.list_projects()
for project in projects['projects']:
    print(f"{project['name']}: {project['tasks_count']} tasks")

Get project details:

project = client.get_project(project_id="uuid-here")
print(f"Project: {project['name']}")
print(f"Features: {len(project['features'])}")
print(f"Tasks: {len(project['tasks'])}")

Create new project:

result = client.create_project(
    name="API Redesign",
    description="Complete API overhaul with v2 endpoints"
)
project_id = result['project']['id']

5. Task Management

Create tasks:

from scripts.archon_client import ArchonClient

# Use the host URL provided by user
archon_host = "http://localhost:8181"  # Replace with user's actual host
client = ArchonClient(base_url=archon_host)

task = client.create_task(
    project_id="project-uuid",
    title="Implement OAuth2 authentication",
    description="Add OAuth2 flow with JWT tokens",
    status="todo"  # todo, in_progress, done, blocked
)

Update task status:

client.update_task(
    task_id="task-uuid",
    updates={"status": "in_progress"}
)

List and filter tasks:

# Get all in-progress tasks for a project
tasks = client.list_tasks(
    project_id="project-uuid",
    status="in_progress",
    limit=20
)

# Get task details
task = client.get_task(task_id="task-uuid")

Task statuses:

• "todo": Not started
• "in_progress": Currently working
• "done": Completed
• "blocked": Blocked by dependencies

6. Document Management

Create versioned documents:

from scripts.archon_client import ArchonClient

# Use the host URL provided by user
archon_host = "http://localhost:8181"  # Replace with user's actual host
client = ArchonClient(base_url=archon_host)

doc = client.create_document(
    title="API Specification",
    content="# API Spec\n\nDetailed specification...",
    project_id="project-uuid"  # Optional
)

Update documents (automatic versioning):

client.update_document(
    document_id="doc-uuid",
    updates={
        "title": "Updated API Spec",
        "content": "# Updated Spec\n\nNew content..."
    }
)

List documents:

# All documents
docs = client.list_documents()

# Project-specific documents
docs = client.list_documents(project_id="project-uuid")

Common Workflows

Note: All workflows below assume you've already obtained the Archon host URL from the user and verified the connection. Use that URL when creating the ArchonClient.

Search-First Workflow

Always search Archon before other sources:

from scripts.archon_client import ArchonClient

# Use the host URL provided by user earlier in conversation
archon_host = "http://localhost:8181"  # Replace with user's actual host
client = ArchonClient(base_url=archon_host)

# 1. Search Archon first
results = client.search_knowledge("Next.js API routes", top_k=5)

if results.get('results'):
    # Found in Archon - use this knowledge
    for result in results['results']:
        print(result['content'])
else:
    # Not in Archon - could crawl documentation
    print("No results in Archon. Consider crawling Next.js docs:")
    client.crawl_website("https://nextjs.org/docs")

Project Setup Workflow

Setting up a new development project:

from scripts.archon_client import ArchonClient

# Use the host URL provided by user
archon_host = "http://localhost:8181"  # Replace with user's actual host
client = ArchonClient(base_url=archon_host)

# 1. Create project
project = client.create_project(
    name="User Authentication System",
    description="Implement secure user authentication"
)
project_id = project['project']['id']

# 2. Create initial tasks
tasks = [
    "Research authentication libraries",
    "Design database schema",
    "Implement login endpoint",
    "Add JWT token generation",
    "Create password reset flow"
]

for task_title in tasks:
    client.create_task(
        project_id=project_id,
        title=task_title,
        status="todo"
    )

# 3. Search for implementation guidance
results = client.search_knowledge("JWT authentication best practices", top_k=10)

Documentation Indexing Workflow

Building a searchable knowledge base:

from scripts.archon_client import ArchonClient

# Use the host URL provided by user
archon_host = "http://localhost:8181"  # Replace with user's actual host
client = ArchonClient(base_url=archon_host)

# 1. Crawl primary documentation
client.crawl_website("https://docs.framework.com", crawl_depth=3)

# 2. Upload additional resources
client.upload_document(
    "/path/to/internal-guide.pdf",
    metadata={"source_type": "pdf", "tags": ["internal", "guide"]}
)

# 3. Search across all indexed content
results = client.search_knowledge("deployment configuration", top_k=10)

Error Handling

All API calls return standard response format:

Success:

{
  "success": true,
  "data": { /* response payload */ }
}

Error:

{
  "success": false,
  "error": {
    "code": "VALIDATION_ERROR",
    "message": "Invalid parameters"
  }
}

Check for errors:

result = client.search_knowledge("query")
if not result.get('success', True):
    print(f"Error: {result['error']['message']}")

Resources

scripts/archon_client.py

Complete Python client for all Archon API endpoints. Provides the ArchonClient class with methods for:

• Knowledge search and management
• Project and task operations
• Document versioning
• Website crawling
• Standardized error handling

Import and use with user-provided host:

import sys
sys.path.insert(0, '.claude/skills/archon/scripts')
from archon_client import ArchonClient

# Always use the host URL obtained from the user
archon_host = "http://localhost:8181"  # Replace with user's actual host
client = ArchonClient(base_url=archon_host)

scripts/list_knowledge.py

Helper script to quickly list all knowledge base items with connection verification.

Usage:

cd .claude/skills/archon/scripts
python3 list_knowledge.py                      # Uses default localhost:8181
python3 list_knowledge.py http://192.168.1.100:8181  # Custom host

Output:

• Connection status
• Total knowledge items count
• Items grouped by source type
• Detailed list with titles, types, chunks, and source URLs

references/api_reference.md

MANDATORY READING - Complete REST API documentation with authoritative endpoint paths.

ALWAYS read this FIRST before any API operations.

This document contains:

• Correct endpoint paths (e.g., /api/knowledge-items/search, NOT /api/knowledge/search)
• Request/response formats with exact field names
• Query parameter specifications
• Error handling patterns
• All 14 MCP-equivalent endpoints

Read this when:

• Starting any Archon task (MANDATORY)
• Making direct API calls
• Debugging API errors (404s, 400s)
• Verifying Python client implementations
• Understanding request/response formats

Configuration

Host URL: Provided by user at skill activation (e.g., http://localhost:8181, http://192.168.1.100:8181)

Default Settings:

• Default search: hybrid strategy with reranking
• Default crawl depth: 3 levels
• Default results: 10 items

Using Custom Host:

from scripts.archon_client import ArchonClient

# Always use the host URL provided by the user
archon_host = "http://192.168.1.100:8181"  # Example
client = ArchonClient(base_url=archon_host)

Archon Environment Variables (configured on Archon server):

ARCHON_SERVER_PORT=8181  # API server port
SUPABASE_URL=https://your-project.supabase.co
SUPABASE_SERVICE_KEY=your-key
OPENAI_API_KEY=your-key  # For embeddings

Limitations

• Network access required: Archon must be accessible at the provided host URL
• Rate limits: Subject to OpenAI rate limits for embeddings (configured on Archon server)
• Context length: Large documents automatically chunked by Archon
• Crawl depth: Maximum depth of 5 levels
• File size: Practical limit ~100MB per upload