fix: use responses API (backport #83)

[mergify]

Use the OpenAI Responses API for model calls and upgrade langchain-openai, avoiding model-specific handling for reasoning models with tools. Extract output through AIMessage.text so structured reasoning metadata never appears in chat responses.

---

This is an automatic backport of pull request #83 done by Mergify.

[greptile-apps]

Confidence Score: 5/5

The PR appears safe to merge, with the intentional Responses API compatibility requirement clearly documented.

The dependency upgrade accompanies the new model option, both model-call paths use the same output extraction contract, and the changed tests cover Responses API activation and text-only rendering without revealing reasoning metadata.

Sequence Diagram

sequenceDiagram
    participant User
    participant Agent as Ask ALYF Agent
    participant LC as LangChain ChatOpenAI
    participant API as Responses API
    User->>Agent: Send chat or field-writing request
    Agent->>LC: Invoke configured model
    LC->>API: POST /v1/responses
    API-->>LC: Reasoning and text content blocks
    LC-->>Agent: AIMessage
    Agent->>Agent: Read AIMessage.text
    Agent-->>User: Final answer text only

Loading

_{Reviews (1): Last reviewed commit: "fix: use responses API (#83)" | Re-trigger Greptile}