Examples

These examples show how to use json_repair in common integration points without changing the library itself.

Quick start

Run the standard-library-only examples with:

uv run python examples/repair_llm_output.py
uv run python examples/chinese_llm_output.py
uv run python examples/pydantic_schema.py
uv run python examples/stream_stable.py

The FastAPI example needs extra dependencies:

uv add --group dev fastapi uvicorn
uv run uvicorn examples.fastapi_app:app --reload

Included examples

• repair_llm_output.py: Repair JSON wrapped in markdown fences, comments, or extra prose from an LLM response.
• chinese_llm_output.py: Repair Chinese-language JSON while preserving non-Latin characters in the final output.
• pydantic_schema.py: Use a Pydantic v2 model as schema guidance, then validate the repaired object.
• stream_stable.py: Keep a stable best-effort JSON snapshot while a streamed response is still incomplete.
• fastapi_app.py: Repair and validate model output inside a FastAPI endpoint before returning a typed response.