pi-free 2.2.6 → 2.2.8

package/CHANGELOG.md

@@ -7,17 +7,44 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+## [2.2.8] - 2026-07-19
+
+### Changed
+
+- **Pi 0.80.8 compatibility** — migrate OAuth checks to Pi's runtime APIs and raise the peer dependency floor to `>=0.80.8`.
+- **Remove deprecated providers** — remove the Qwen and Modal providers now that their supported free tiers are no longer available.
+- **CI hardening** — pin GitHub Actions to immutable commit SHAs.
+
+### Fixed
+
+- **Pi runtime auth compatibility** — guard OAuth detection when the model registry compatibility APIs are unavailable.
+
+## [2.2.7] - 2026-07-10
+
+### Added
+
+- **AnyAPI provider** — add AnyAPI's OpenAI-compatible gateway with dynamic model discovery, free-model filtering, a 100K-token daily free plan, and `/toggle-anyapi` for switching to the full catalog.
+
+### Changed
+
+- **Pi 0.80 compatibility** — update the Pi SDK peer requirements to `>=0.80.0` and refresh the development toolchain dependencies.
+
+### Fixed
+
+- **AnyAPI model limits** — enrich context and output limits from models.dev when AnyAPI omits them, avoiding incorrect 4K defaults and migrating existing caches once.
+- **TokenRouter Pi AI compatibility** — use the supported lazy OpenAI completions API path for Pi 0.80.
+
 ## [2.2.6] - 2026-07-09
 
 ### Changed
 
-- **Faster startup (~30x)** — warm-cache load dropped from ~2.1s to ~70ms: providers now serve from a 1-hour disk cache and fetch live only on cold/stale cache. Extends the cache-first pattern (already used by Cline) to kilo, fastrouter, and all OpenAI-compatible fetchers (tokenrouter, zenmux, crofai, deepinfra, sambanova, together, novita, routeway, bai, openmodel); the dynamic built-in phase runs concurrently with the static providers.
-- **Coding-Index debug logging is now opt-in** — `~/.pi/modelmatch.log` previously received one synchronous `appendFileSync` per model per match attempt at startup. Now off by default; set `PI_FREE_BENCHMARK_DEBUG=1` to re-enable.
-- **Config reads are memoized** — `~/.pi/free.json` is parsed once and reused while its mtime is unchanged.
+- **Faster startup (~30x)** — warm-cache load dropped from ~2.1s to ~70ms: providers now serve from a 1-hour disk cache and fetch live only on cold/stale cache. Extends the cache-first pattern (already used by Cline) to kilo, fastrouter, and all OpenAI-compatible fetchers (tokenrouter, zenmux, crofai, deepinfra, sambanova, together, novita, routeway, bai, openmodel); the dynamic built-in phase runs concurrently with the static providers. Thanks @lmilojevicc
+- **Coding-Index debug logging is now opt-in** — `~/.pi/modelmatch.log` previously received one synchronous `appendFileSync` per model per match attempt at startup. Now off by default; set `PI_FREE_BENCHMARK_DEBUG=1` to re-enable. Thanks @lmilojevicc
+- **Config reads are memoized** — `~/.pi/free.json` is parsed once and reused while its mtime is unchanged. Thanks @lmilojevicc
 
 ### Fixed
 
-- **Cache-poisoning guard** — a transient partial API response (a 200 returning a near-empty list) can no longer overwrite a healthy cached model list; fetches returning < 50% of the cached count keep the existing cache.
+- **Cache-poisoning guard** — a transient partial API response (a 200 returning a near-empty list) can no longer overwrite a healthy cached model list; fetches returning < 50% of the cached count keep the existing cache. Thanks @lmilojevicc
 
 ### Removed
 

package/README.md

@@ -14,7 +14,7 @@ Free and paid AI model providers for [Pi](https://pi.dev). Access **free and pai
 
 When you install pi-free, it:
 
-1. **Registers free-tier providers** — Kilo, Cline, LLM7, OpenModel, TokenRouter, and more
+1. **Registers free-tier providers** — Kilo, Cline, LLM7, OpenModel, TokenRouter, AnyAPI, and more
 2. **Captures Pi's built-in providers** with free/paid toggles — OpenCode, OpenRouter
 3. **Fetches models dynamically** — ZenMux, CrofAI, Mistral, Groq, Cerebras, xAI, Hugging Face when API keys are configured
 4. **Filters to show only free models by default** — paid models hidden until explicitly toggled on
@@ -62,6 +62,7 @@ First run creates `~/.pi/free.json` automatically. Add keys there or use environ
 ```json
 {
   "ollama_api_key": "...",
+  "anyapi_api_key": "...",
   "sambanova_api_key": "...",
   "deepinfra_api_key": "..."
 }
@@ -74,7 +75,7 @@ First run creates `~/.pi/free.json` automatically. Add keys there or use environ
 | Category | Providers |
 |---|---|
 | ✅ **Free** | Kilo, Cline, OpenRouter, OpenCode, LLM7, OpenModel, TokenRouter (1 free) |
-| 🔄 **Freemium** | Ollama Cloud, SambaNova, AgentRouter |
+| 🔄 **Freemium** | AnyAPI, Ollama Cloud, SambaNova |
 | 💳 **Paid** | ZenMux, CrofAI, DeepInfra, Together, Novita, Routeway, b.ai |
 | 🔧 **Dynamic** | Mistral, Groq, Cerebras, xAI, Hugging Face, FastRouter |