No description
- TypeScript 99.7%
- Shell 0.3%
Inject glm-5.2 into the opencode-go provider via ensureIluPiModelsJson, matching Z.AI official specs: 1M context window, 131072 max output tokens, reasoning-only with high/max effort levels (off is unsupported). Refactor pi-models.ts from single-model injection (qwen3.7-max only) to a loop over OPENCODE_GO_INJECTED_MODELS so future models can be added by appending to the array. The stale-entry check now uses full deep equality to catch drift in any field, not just api/baseUrl. Update agent-models.ts to register glm-5.2 in the picker at the correct position with its 1M context window constant. Update ui-controller and pi-adapter tests to cover the new model in selector lists, context metadata, and registry resolution. Expand pi-models tests to cover both injected models across all scenarios (first-run, idempotency, stale update, preservation, corrupt JSON, missing providers, non-object provider, and Pi registry resolution). Verified against OpenCode upstream (anomalyco/opencode issues #32172 and #32444) and the Pi SDK's built-in opencode-go model definitions. |
||
|---|---|---|
| .pi | ||
| bin | ||
| docs | ||
| scripts | ||
| src | ||
| test | ||
| .gitignore | ||
| .prettierrc.json | ||
| AGENTS.md | ||
| package-lock.json | ||
| package.json | ||
| tsconfig.json | ||