chore(examples): bump example models to GPT-6.2 Sol, Claude Haiku 5.5 and Sonnet 5.5
Moves the examples and README snippets off gpt-5.1 and the Claude 4.5 models onto current ones. Each changed example was run against the live model APIs to confirm it still does the full tool-call loop.
| Example | Before | After |
|---|---|---|
openai-responses-integration.ts | gpt-5.1 | gpt-6.2-sol |
ai-sdk-integration.ts | gpt-5.1 | gpt-6.2-sol |
anthropic-integration.ts | claude-haiku-4-5-20251001 | claude-haiku-5-5 |
claude-agent-sdk-integration.ts | claude-sonnet-4-5-20250929 | claude-sonnet-5-5 |
openai-integration.ts (Chat Completions) | gpt-5.1 | gpt-6.1-terra + reasoning_effort: 'none' |
Why Chat Completions is on 6.1 Terra. The GPT-6.2 models reject function tools on /v1/chat/completions (Function tools with reasoning_effort are not supported for gpt-6.2-sol in /v1/chat/completions. To use function tools, use /v1/responses or set reasoning_effort to 'none'), and they also reject 'none'. gpt-6.1-terra and gpt-6-luna accept tools on Chat Completions with reasoning_effort: 'none', so the example uses Terra and points to the Responses example for tools plus reasoning.
Verification
Each example was run with tsx against the live model API. StackOne /accounts and /mcp were served by the repo's MSW mock, with an account that has workday_* tools.
| Example | Model served | Result |
|---|---|---|
| openai-integration | gpt-6.1-terra | tool call workday_list_workers, tool result sent back, final answer |
| openai-responses-integration | gpt-6.2-sol | function call returned |
| ai-sdk-integration | gpt-6.2-sol | 2-step loop, tool executed, final answer |
| anthropic-integration | claude-haiku-5.5 | tool_use blocks returned |
| claude-agent-sdk-integration | claude-sonnet-5.5 | both MCP tools called, final answer |
oxfmt --check is clean, oxlint --type-aware --type-check reports 0 warnings, and vitest run passes (1319 passed, 5 skipped because they need dist/).
Requested by guillaume@stackone.com
Post-merge validation
No runtime change to the package: this touches examples, README and one JSDoc line. The repo's tests only run the examples against mocks, so the signal after merge is a manual pnpm run:example examples/<file>.ts with real keys.
š¤ Generated with Claude Code
Summary by cubic
Updates the examples and README snippets from gpt-5.1 and the Claude 4.5 models to current ones (gpt-6.2-sol, claude-haiku-5-5, claude-sonnet-5-5), plus the matching JSDoc comment in src/tool.ts.
- Chat Completions now uses
gpt-6.1-terrawithreasoning_effort: 'none'because the GPT-6.2 models reject function tools on/v1/chat/completionsand don't accept'none'either; the Responses API supports both, so that example points there. - Each example was run against the live model API and the full tool-call loop was verified.
- No runtime code changes;
vitest runpasses and linting is clean.
Review gates and proof
Recorded by s2d2 on d697624.
| gate | result |
|---|---|
| forge | not run |
| security_scan | not run |
| codex_security | not run |
| security_review | not run |
| code_review | not run |
| demo | ran |
Proof is in the requester's thread: this repository is public.