deadends-dev 0.3.3

Structured failure knowledge for AI agents — dead ends, workarounds, error chains

deadends.dev

Structured failure knowledge for AI coding agents.

970+ error entries across 20 domains. When AI coding agents encounter errors, they waste tokens on approaches that are known to fail. deadends.dev tells agents what NOT to try, what actually works, and what error comes next.

Website: deadends.dev · MCP Server: Smithery · PyPI: deadends-dev · API: /api/v1/index.json

Installation

pip install deadends-dev

Requirements: Python 3.10+

MCP Server

The MCP server exposes 8 tools for AI coding agents:

Tool	Description
lookup_error	Match an error message against 970+ known patterns. Returns dead ends, workarounds, and error chains.
get_error_detail	Get full details for a specific error by ID (e.g., python/modulenotfounderror/py311-linux).
list_error_domains	List all 20 error domains and their counts.
search_errors	Fuzzy keyword search across all domains (e.g., "memory limit", "permission denied").
list_errors_by_domain	List all errors in a specific domain, sorted by fix rate, name, or confidence.
batch_lookup	Look up multiple error messages at once (max 10).
get_domain_stats	Get quality metrics for a domain: avg fix rate, resolvability, confidence breakdown.
get_error_chain	Traverse the error transition graph: what errors follow, precede, or get confused with this one.

Local (Claude Desktop / Cursor)

Add to ~/.claude/claude_desktop_config.json:

{
  "mcpServers": {
    "deadend": {
      "command": "python",
      "args": ["-m", "mcp.server"],
      "cwd": "/path/to/deadends.dev"
    }
  }
}

Hosted (Smithery — no local setup)

Install via Smithery:

# Claude Code
npx -y @smithery/cli@latest install deadend/deadends-dev --client claude

# Cursor
npx -y @smithery/cli@latest install deadend/deadends-dev --client cursor

Or connect directly: https://server.smithery.ai/deadend/deadends-dev

Example Response

When an agent encounters ModuleNotFoundError: No module named 'torch', the lookup_error tool returns:

## ModuleNotFoundError: No module named 'X' (Python 3.11+)
Resolvable: true | Fix rate: 0.88

### Dead Ends (DO NOT TRY):
- pip install X with system Python (fails 70%): venv not activated

### Workarounds (TRY THESE):
- Create venv, activate, then pip install (works 95%)
- Use python -m pip install instead of bare pip (works 90%)

Quick Start — Python SDK

from generator.lookup import lookup, batch_lookup, search

# Single error lookup
result = lookup("ModuleNotFoundError: No module named 'torch'")

# What NOT to try (saves tokens and time)
for d in result["dead_ends"]:
    print(f"AVOID: {d['action']} — fails {int(d['fail_rate']*100)}%")

# What actually works
for w in result["workarounds"]:
    print(f"TRY: {w['action']} — works {int(w['success_rate']*100)}%")

# Batch lookup (multiple errors at once)
results = batch_lookup([
    "ModuleNotFoundError: No module named 'torch'",
    "CUDA error: out of memory",
    "CrashLoopBackOff",
])

# Keyword search
hits = search("memory limit", domain="docker", limit=5)

Quick Start — CLI

pip install deadends-dev
deadends "CUDA error: out of memory"
deadends --list  # show all known errors

API Endpoints

Endpoint	Description
/api/v1/match.json	Lightweight regex matching (fits in context window)
/api/v1/index.json	Full error index with all metadata
/api/v1/{domain}/{slug}/{env}.json	Individual ErrorCanon (example)
/api/v1/openapi.json	OpenAPI 3.1 spec with response examples
/api/v1/stats.json	Dataset quality metrics by domain
/api/v1/errors.ndjson	NDJSON streaming (one error per line)
/api/v1/version.json	Service metadata and endpoint directory
/llms.txt	LLM-optimized error listing (llmstxt.org standard)
/llms-full.txt	Complete database dump
/.well-known/ai-plugin.json	AI plugin manifest
/.well-known/agent-card.json	Google A2A agent card
/.well-known/security.txt	Security contact (RFC 9116)

Covered Domains (20)

Domain	Errors	Examples
Python	86	ModuleNotFoundError, TypeError, KeyError, MemoryError, RecursionError
Node	70	ERR_MODULE_NOT_FOUND, EACCES, EADDRINUSE, heap OOM, ERR_REQUIRE_ESM
Kubernetes	55	CrashLoopBackOff, ImagePullBackOff, OOMKilled, RBAC forbidden, HPA
Docker	55	no space left, exec format error, bind address in use, healthcheck
AWS	50	AccessDenied, S3 NoSuchBucket, Lambda timeout, CloudFormation rollback
Java	50	NullPointerException, ClassNotFound, OutOfMemoryError, connection pool
Go	50	nil pointer, unused import, interface conversion, slice out of range
Database	50	deadlock, connection pool, slow query, replication lag, constraint violation
Git	47	failed to push, merge conflicts, detached HEAD, stash apply, tags
TypeScript	45	TS2307, TS2322, TS2345, TS2532, TS7053, TS2769, TS18048
Terraform	45	state lock, cycle, provider not found, moved block, backend init
Rust	45	E0382 borrow, E0308 mismatch, E0277 trait, E0106 lifetime, E0507
React	45	invalid hook call, too many re-renders, unique key, context, act()
CUDA	45	OOM, device-side assert, NCCL, cuDNN, tensor device mismatch
CI/CD	44	GitHub Actions timeout, secret not found, Docker rate limit, cache miss
PHP	43	headers already sent, too many connections, autoload, memory exhaustion
pip	40	build wheel failed, conflicting deps, externally-managed, hash mismatch
Networking	40	connection refused, ECONNRESET, SSL certificate, DNS timeout, EPIPE
Next.js	35	hydration failed, dynamic server, server-only import, RSC serialization
.NET	28	NullReferenceException, LINQ translation, DI circular, EF concurrency

ErrorCanon Data Format

Each error is a JSON file with:

{
  "error": { "signature": "...", "regex": "...", "domain": "..." },
  "verdict": { "resolvable": "true|partial|false", "fix_success_rate": 0.88 },
  "dead_ends": [{ "action": "...", "why_fails": "...", "fail_rate": 0.75 }],
  "workarounds": [{ "action": "...", "success_rate": 0.92, "how": "..." }],
  "transition_graph": { "leads_to": [...], "preceded_by": [...] }
}

AI Coding Agent Integration — 18 Discovery Formats

Every page on deadends.dev includes machine-readable data in 18 formats:

Format	Location	Purpose
JSON API	/api/v1/{id}.json	RESTful error data per ErrorCanon
match.json	/api/v1/match.json	Compact regex-only file (load entire DB into context)
index.json	/api/v1/index.json	Master error index with metadata
stats.json	/api/v1/stats.json	Dataset quality metrics per domain
errors.ndjson	/api/v1/errors.ndjson	Streaming NDJSON for batch processing
OpenAPI	/api/v1/openapi.json	Full API spec with response examples
JSON-LD	Every <head>	Schema.org TechArticle + FAQPage
ai-summary	Every page	<pre id="ai-summary"> KEY=VALUE blocks
llms.txt	/llms.txt	llmstxt.org standard
llms-full.txt	/llms-full.txt	Complete database dump
ai-plugin.json	/.well-known/	OpenAI plugin manifest
agent-card.json	/.well-known/	Google A2A protocol
security.txt	/.well-known/	RFC 9116 security contact
robots.txt	/robots.txt	34 AI crawlers explicitly welcomed
CLAUDE.md	/CLAUDE.md	Claude Code instructions
AGENTS.md	/AGENTS.md	OpenAI Codex CLI instructions
.clinerules	/.clinerules	Cline AI instructions

Development

pip install -e ".[dev]"

# Full pipeline (validate → generate → build → test)
python -m generator.pipeline

# Individual steps
python -m generator.bulk_generate     # Generate canons from seeds
python -m generator.build_site        # Build static site
python -m generator.validate          # Validate data + site
python -m pytest tests/ -v            # Run tests

Contributing

Add error definitions to generator/bulk_generate.py or create JSON files directly in data/canons/.

python -m generator.validate --data-only  # Validate before submitting

License

MIT (code) · CC BY 4.0 (data)