# hinthornw/trustcall

Tenacious tool calling built on LangGraph

Repository: https://github.com/hinthornw/trustcall
Canonical: https://ross.abutalabs.com/products/trustcall
Language: Python
License: MIT
License Family: permissive
Last push: 2025-07-17T20:57:10+00:00

## Health v2 (maintenance only)
Score: 38/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 32, release rhythm 35, longevity 57
- inputs: {"age_days": 799, "days_push": 412, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1089, forks 89 (observed 2026-08-28T04:03:32.638190+00:00)

## What it is
A Python library that improves LLM tool calling reliability by having models generate JSON patch operations instead of full JSON blobs. It provides resilient validation retries and accurate schema updates for complex, nested schemas built on LangGraph.

## Use cases
- extract structured data from text with llms
- populate complex nested pydantic schemas reliably
- update existing  schemas without losing data
- retry llm validation errors automatically
- build multi-step agent tool use pipelines
- route llm outputs to tools

## When to choose
- you need reliable structured output from llms for complex nested schemas
- you want to update existing extraction results without information loss
- your stack already uses LangChain or LangGraph

## When to avoid
- you need a language other than Python
- you don't use LLM tool calling or structured extraction
- you want a provider-agnostic solution without LangGraph dependencies

## Facets
- artifact type: library
- maturity: active
- function: llm-inference, agent-framework, serialization, machine-learning
- domain: large-language-models, artificial-intelligence, developer-tools
- platform: python
- tags: tool-calling, structured-output, -patch, pydantic, langgraph, extraction, ai-agents

## Member repositories
- hinthornw/trustcall (main) score 38

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:32.638190+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T06:49:17.920208+00:00, confidence not recorded.
  - readme: https://github.com/hinthornw/trustcall (fetched 2026-08-28T04:03:32.638190+00:00, sha 7c8b01e796ab)
- Data as of 2026-08-30T08:39:29.467469+00:00.