# CHANGELOG

All notable changes to this project will be documented in this file.

## [0.6.0] - 2026-06-05

- Fix `apiKey` registered as a literal string instead of an environment variable reference. Changed `apiKey: "OLLAMA_API_KEY"` to `apiKey: "$OLLAMA_API_KEY"` in `registerProvider`, resolving the deprecation warning emitted by pi v0.77.0+ and making the `OLLAMA_API_KEY` env var work alongside `auth.json` (env var takes priority, falls back to `auth.json`). Thanks @mandusm (#21).
- Make `scripts/generate-models.ts` produce a stable output: sort models by `id` and each object's keys alphabetically (with `id` and `name` first), so regenerating the catalog produces minimal diffs. Thanks @shyim for the idea (#22).
- Exclude models announced for retirement (effective 2026-06-16) from the generated `GENERATED_MODELS` list: `kimi-k2-thinking`, `kimi-k2:1t`, `minimax-m2`, `glm-4.6`, `qwen3-next:80b`, `qwen3-vl:235b`, `qwen3-vl:235b-instruct`, `cogito-2.1:671b`. See https://docs.ollama.com/cloud#deprecations. `/ollama-cloud-refresh` will still register them from the live API until Ollama removes them.

## [0.5.0] - 2026-05-22

- Update `peerDependencies` and imports to the new `@earendil-works/*` packages.
- Add build script (`scripts/generate-models.ts`) to generate `models.generated.ts` with Ollama Cloud tool-capable models from live API data. The script requires no API key.
- Drop `FALLBACK_MODELS` in favor of `GENERATED_MODELS` which ships with the package.
- Remove `apiKey` requirement from `fetchModelIds` and `fetchModelDetails`. Ollama Cloud's public `/v1/models` and `/api/show` endpoints do not require authentication for model metadata. (#15)
- Remove `getOllamaCloudApiKey`, `refreshModelsFromAuth`, and the "No API key found" notification path from `fetchModels`. Users can now browse model metadata without configuring a key.
- Remove `401`/`403` error handling from the model metadata fetch functions.
- Add configuration system (`config.ts`) with JSON config files (global + project-local) and `/ollama-webtools` command for runtime web tools toggling.
- Add test infrastructure (`vitest`) and model validation tests (`test/models.test.ts`).
- Add HTTP status-aware error handling in web tools (`401`/`403` → auth error, `429` → rate limit).
- Add explicit `buildCompat()` with all 17 `OpenAICompletionsCompat` flags set explicitly, verified against Ollama API docs.
- Set the provider display name to `Ollama Cloud` so it reads cleanly in `/login`, `/model`, and the model selector.
- Document `/login` as the recommended way to configure the API key.
- Update README to clarify baked-in model list and refresh behavior.
- Document web-tools.ts module boundaries in file header comment.

## [0.4.1] - 2026-05-07

- Add `renderCall` to `ollama_web_search` and `ollama_web_fetch` tools so the TUI displays the query/URL in the tool call header instead of just the bare tool name. (#12)

## [0.4.0] - 2026-05-06

- Fix `/api/chat` requests not disabling thinking when Pi's thinking level is set to `off`. Maps Pi `off` to `reasoning_effort: "none"` on models where the API respects it, hides the `off` level on models where it doesn't (gpt-oss, kimi-k2-thinking, minimax, qwen3-vl). (#6)
- Add `thinking-levels.ts` with curated per-model thinking level maps (DEFAULT, GPT_OSS, QWEN3, NO_OFF), validated against all 24 thinking-capable models via automated experiment (see docs/think-experiment.md).
- Fix system prompt (AGENTS.md content) not being read by GLM models by setting `supportsDeveloperRole: false` on all registered models.
- Rename smoke test to test, add lint step.
- Treat stale local model caches as usable for immediate startup while triggering the same visible refresh flow as `/ollama-cloud-refresh` on `session_start`; use fallback models only when the cache is missing or invalid.
- Add a single-line `/ollama-cloud-refresh` progress widget showing the current stage, count, percentage, failures, and progress bar.
- Add thinking on/off assertions to the CI test workflow.

## [0.3.1] - 2026-05-05

- Fix `OLLAMA_API_KEY` env var not being respected by `fetchModels` and web tools. pi-ai does not know about the `ollama-cloud` provider ID, so `AuthStorage.getApiKey()` alone misses the env var. Added explicit `process.env.OLLAMA_API_KEY` fallback.
- Switch web tools to `AuthStorage.create()` for API key lookup, matching the `models.ts` auth pattern from v0.2.1.
- Add null-safe access to `data.details?.family` in `resolveThinkingLevelMap`.
- Change `OLLAMA_BASE` from `export let` to `export const` to prevent accidental mutation.
- Fix fallback model IDs to use real Ollama Cloud identifiers (`glm-5.1`, `gemma4:31b`) instead of synthetic `:cloud` suffixes.
- Add smoke test workflow for CI.

## [0.3.0] - 2026-05-04

- Derive `thinkingLevelMap` from pi's built-in model definitions instead of hardcoding model-family mappings. The extension now picks up thinking level metadata automatically when pi-mono adds or updates it for any model.
- Add family-based fallback matching: when an Ollama Cloud model ID doesn't match a pi model ID exactly, the extension now tries matching by model family (via Ollama's `details.family` field). For example, `gemma4:31b` correctly picks up Gemma 4's thinking level map from pi.

## [0.2.1] - 2026-04-29

- Fix API key retrieval by using `AuthStorage` instead of `ctx.modelRegistry.getApiKeyForProvider`. The provider-level API key lookup was failing, causing auth to only work when an environment variable was set. Now reads from `auth.json` directly via the pi `AuthStorage` class.

## [0.2.0] - 2026-04-28

- Add `PI_OLLAMA_WEB_TOOLS` environment variable to optionally disable `ollama_web_search` and `ollama_web_fetch` tool registrations. Set to `0`, `false`, `no`, `off`, or an empty string to opt-out. The model provider and `/ollama-cloud-refresh` command remain active regardless.