Doc-Squeeze — ClawHub Skill

Overview

Doc-Squeeze lets AI agents read external documentation and extract structured knowledge without a browser. It fetches any URL as clean markdown, extracts structured JSON from schema definitions, searches within documents, and processes multiple URLs in parallel.

Cost: $0 — Both Jina Reader and Groq have generous free tiers.

Tools

squeeze_url

Property	Value
Endpoint	POST /api/squeeze
Auth	Optional (API key for higher limits)
Latency	~2-5s (fetch) + ~1-3s (filter)

Input:

{
  "url": "https://docs.stripe.com/api/authentication",
  "focus": "Python code for setting the API key"
}

Field	Type	Required	Description
url	string	✅	URL of the documentation page.
focus	string	❌	Topic filter — triggers LLM.

---

extract_structured

Property	Value
Endpoint	POST /api/extract
Auth	Optional
Latency	~3-8s

Input:

{
  "url": "https://docs.stripe.com/api",
  "schema_definition": {
    "endpoints": [{"method": "str", "path": "str", "description": "str"}],
    "auth_methods": ["str"]
  },
  "instructions": "Focus on the payments API only"
}

Field	Type	Required	Description
url	string	✅	URL to extract from.
schema_definition	object	✅	JSON schema defining what to extract.
instructions	string	❌	Additional extraction guidance.

---

search_docs

Property	Value
Endpoint	POST /api/search
Auth	Optional
Latency	~3-8s

Input:

{
  "url": "https://docs.python.org/3/library/asyncio.html",
  "query": "How do I run multiple coroutines concurrently?",
  "max_results": 3
}

Field	Type	Required	Description
url	string	✅	URL to search within.
query	string	✅	What to search for.
max_results	integer	❌	Number of results (1-10, default 3).

---

batch_squeeze

Property	Value
Endpoint	POST /api/batch
Auth	Optional
Latency	~3-15s (parallel)

Input:

{
  "urls": [
    "https://docs.stripe.com/api/authentication",
    "https://docs.stripe.com/api/errors"
  ],
  "focus": "error handling"
}

Field	Type	Required	Description
urls	array	✅	URLs to fetch (max 10).
focus	string	❌	Optional topic filter for all URLs.

---

Self-Discovery

Agents can introspect the full tool schema at runtime:

GET /api/skill          → openclaw.json manifest
GET /.well-known/mcp.json → MCP server discovery
GET /.well-known/ai-plugin.json → OpenAI plugin manifest

Authentication

Tier	Rate Limit	How to Get
Free	5/minute	No key needed
Dev	60/minute	POST /api/keys/create
Pro	300/minute	Contact us

API keys are passed via X-API-Key header.

Permissions

Permission	Host	Required	Reason
Network Access	r.jina.ai	✅	Fetches docs as markdown.
Network Access	api.groq.com	❌	LLM filtering (only with focus)

Environment

Variable	Required	How to get it
GROQ_API_KEY	No*	Free at console.groq.com/keys

*Without the key, only raw squeeze and batch work. Extract and search require Groq.

Agent Usage Example

import requests

BASE = "https://doc-squeeze.onrender.com"

# 1. Raw fetch
resp = requests.post(f"{BASE}/api/squeeze", json={
    "url": "https://docs.python.org/3/library/json.html"
})
docs = resp.json()["markdown"]

# 2. Structured extraction
resp = requests.post(f"{BASE}/api/extract", json={
    "url": "https://docs.stripe.com/api",
    "schema_definition": {
        "endpoints": [{"method": "str", "path": "str"}],
        "auth_type": "str"
    }
})
data = resp.json()["data"]

# 3. Deep search
resp = requests.post(f"{BASE}/api/search", json={
    "url": "https://docs.python.org/3/library/asyncio.html",
    "query": "How to cancel a task?"
})
answers = resp.json()["results"]

# 4. Batch fetch
resp = requests.post(f"{BASE}/api/batch", json={
    "urls": ["https://example.com", "https://httpbin.org/html"],
    "focus": "main content"
})
results = resp.json()["results"]