free-coding-models 0.5.45 → 0.5.47

package/README.md

@@ -468,7 +468,7 @@ When a tool mode is active (via `Z`), models incompatible with that tool are hig
 
 ## π Pi Extension — FCM-Pi ⚠️ BETA
 
-**FCM-Pi** is a native [Pi coding agent](https://pi.dev) extension that integrates `free-coding-models` directly into your Pi session. It auto-selects the best free model on startup, shows a live animated scan in the Pi status bar, and lets you hot-swap models mid-session.
+**FCM-Pi** is a native [Pi coding agent](https://pi.dev) extension that integrates `free-coding-models` directly into your Pi session. It stays silent by default, scans only when you run `/fcm`, and lets you explicitly hot-swap models mid-session.
 
 > **BETA** — The extension is under active development and not yet published to npm. Install via local path only.
 >
@@ -492,16 +492,16 @@ Then restart Pi. The extension loads automatically.
 
 | Feature | Description |
 |---------|-------------|
-| **Auto-scan on startup** | On every Pi session start, FCM pings ~30 candidate models in parallel and selects the best one automatically |
-| **Animated progress scan** | While scanning, the Pi status bar shows a live magenta spinner with the model + provider being probed and a real-time % progress |
-| **Provider in status bar** | The active model is displayed as `GLM 4.7 (Cerebras) [FCM S+]` so you always know exactly what's running |
-| **10-minute disk cache** | Results are cached to `~/.pi/agent/fcm-cache.json` — subsequent session starts are instant |
-| **Auto-failover** | If a request fails (HTTP 4xx/5xx), FCM auto-triggers a fresh scan and switches to the next best model |
+| **Silent startup** | No scan, no footer noise, and no automatic model switch on Pi boot or `/resume` |
+| **Manual scan** | `/fcm` pings ~30 candidate models in parallel and waits for your explicit selection |
+| **Temporary progress scan** | The Pi status bar shows live scan progress only while probing/benchmarking, then hides again |
+| **10-minute disk cache** | Results are cached to `~/.pi/agent/fcm-cache.json` for faster diagnostics and lists |
+| **Error-triggered picker** | If a request fails (HTTP 4xx/5xx), FCM reopens the picker and marks the failed model `🔴 BUGGED` |
 | **Daemon integration** | If the FCM router daemon is running (`free-coding-models --daemon-bg`), scan results are fetched instantly from its cache |
 
 ### Scan progress display
 
-During `/fcm` or on session start, the Pi footer status shows:
+During `/fcm`, the Pi footer status shows:
 
 ```
 ⠸ Probing: Kimi K2.6 [Nvidia], Step 3.5 Flash [Stepfun] — 47% (14/30)
@@ -519,7 +519,7 @@ During `/fcm` or on session start, the Pi footer status shows:
 |---------|-------------|
 | `/fcm` | Re-scan and pick a model interactively from the top 10 ranked |
 | `/fcm-list` | Display a ranked table of top 20 available models (SWE / Latency / TPS / Provider) |
-| `/fcm-router` | Connect Pi to the local FCM Smart Router daemon (auto-failover across providers) |
+| `/fcm-router` | Explicitly connect Pi to the local FCM Smart Router daemon |
 | `/fcm-status` | Show diagnostics: active model, last scan source, daemon state |
 
 ### Composite ranking
@@ -535,7 +535,7 @@ Models are ranked using a composite score:
 
 ### Context window limits
 
-Cerebras free-tier API has a strict **8192 total token limit** (prompt + tools + completion). The FCM catalog reflects this accurately so Pi allocates the right budget and doesn't overflow the context silently.
+Cerebras free-tier API has a strict **~8k total token limit** (prompt + tools + completion). Tiny-context models can pass a `hi` probe but fail real agent prompts, so FCM hides 8k Cerebras models from Pi/OpenCode pickers.
 
 ### Full documentation
 
@@ -543,6 +543,35 @@ See [`pi-extension/README.md`](./pi-extension/README.md) for the complete archit
 
 ---
 
+## OpenCode Plugin — fcm-opencode ⚠️ BETA
+
+`fcm-opencode` is the OpenCode adapter for the same FCM scanner/ranker used by FCM-Pi.
+
+### Local install
+
+```bash
+mkdir -p ~/.config/opencode/plugins
+ln -sf /Users/vava/Documents/GitHub/free-coding-models/opencode-plugin/index.js \
+  ~/.config/opencode/plugins/fcm-opencode.js
+```
+
+### Commands
+
+| Command | Description |
+|---------|-------------|
+| `/fcm` | Run an explicit scan and list ranked choices without switching |
+| `/fcm 1` | Switch OpenCode config to ranked model #1 |
+| `/fcm best` | Switch OpenCode config to the best ranked model |
+| `/fcm rescan` | Force a fresh scan |
+| `/fcm status` or `/fcm-status` | Show plugin diagnostics |
+| `/fcm router` or `/fcm-router` | Switch OpenCode config to the local FCM Smart Router daemon |
+
+Startup is intentionally light: fresh cache first, daemon second, **no direct scan** unless you run `/fcm`.
+
+See [`opencode-plugin/README.md`](./opencode-plugin/README.md) for details and limitations.
+
+---
+
 ## ⌨️ TUI Keys
 
 ### Keyboard

package/changelog/v0.5.46.md

@@ -0,0 +1,9 @@
+# Changelog v0.5.46 - 2026-07-08
+
+### Fixed
+- **LLM7 catalog corrected against live `/v1/models` API.** The previous audit had added several models that are NOT free on LLM7. The live API shows that of LLM7's 12 models, only 2 are on the `turbo` (free, `noKeyNeeded`) tier; everything else — including the `claude-*` (Fable 5, Opus 4.8, Sonnet 5), `gpt-5.4` / `gpt-5.4-mini` / `gpt-5.5`, `glm-5.2`, `kimi-k2.6`, `deepseek-v4-flash`, and `minimax-m2.7` entries — are `pro` (usage-based paid) and were incorrectly listed as free.
+- **Removed 10 non-free (pro) models from LLM7**, keeping only the two genuinely-free `turbo` models: `devstral-small-2:24b` (S / 68.0%) and `codestral-latest` (A / 40.0%). Total catalog: 257 → 247 models.
+- Removed the `1.05M` context-window entries (gpt-5.5 / gpt-5.4) that were causing alignment drift; no `1.05M` remains in the catalog.
+
+### Notes
+- LLM7's free (`turbo`) tier now has only 2 coding models. The `llm7` provider remains `noKeyNeeded: true` and should only surface free models per the source-header rule ("Pro-tagged models from /v1/models are intentionally excluded").

package/changelog/v0.5.47.md

@@ -0,0 +1,23 @@
+# Changelog v0.5.47 - 2026-07-08
+
+### Added
+- **New BETA `fcm-opencode` plugin** — an OpenCode adapter that brings `free-coding-models` scans, routing, and provider injection into OpenCode. It reuses the exact same scan / ranking / API-key / daemon / model-safety code as `fcm-pi`, so provider fixes and safety filters stay in one place. Install via a local symlink (see `opencode-plugin/README.md`). Commands:
+  - `/fcm` — scan and list ranked choices **without switching**
+  - `/fcm 1` / `/fcm 2` / … — explicitly switch OpenCode config to ranked model #N
+  - `/fcm best` — explicitly switch to the best ranked model
+  - `/fcm rescan` — force a fresh scan
+  - `/fcm status` (or `/fcm-status`) — plugin diagnostics (cache age, daemon, active model)
+  - `/fcm router` (or `/fcm-router`) — explicitly switch OpenCode to the local FCM Smart Router daemon
+- **New shared `pi-extension/lib/pi-model-config.js` module** — centralizes the Pi model safety policy so both adapters apply identical rules: `parseContextWindow` (`8k` / `128k` / `1M` → number), `getPiMaxTokens` (caps completions to leave prompt room), `isPiContextUsable` (hides tiny-context models that pass a probe but fail real agent prompts, e.g. 8k Cerebras), and `getPiReasoningFlag` (keeps OpenAI-compatible FCM models in non-thinking mode).
+- **`runFcmScan({ mode })` now supports `auto`, `daemon`, and `direct` scan strategies.** OpenCode uses `daemon` at startup so direct probes never block boot.
+
+### Changed
+- **FCM-Pi is now silent by default.** No startup scan, no footer noise, and no automatic model switch on Pi boot or `/resume`. Run `/fcm` to scan and pick a model explicitly.
+- **Error-triggered picker replaces silent auto-failover.** When the active model returns HTTP 4xx/5xx, FCM reopens the picker and marks the failing model `🔴 BUGGED` instead of switching automatically — you stay in control of which model runs.
+- **Pi scan-progress footer now shows the branded `> free-coding-models` badge** in the exact same green/white-on-black colours as the main TUI header logo, instead of scrolling live model names. The live `%` progress and `(completed/total)` counter stay beside it.
+- **OpenCode startup is light by default:** fresh cache first, daemon second, no direct scan unless the user explicitly runs `/fcm`. Long probes never freeze OpenCode boot.
+- **README updated** with a new `fcm-opencode` section and a refreshed FCM-Pi section reflecting the silent-by-default behaviour.
+
+### Notes
+- `fcm-pi` and `fcm-opencode` remain **BETA** and are installed via local path / symlink, not via the npm `free-coding-models` package.
+- The full shared-core extraction into `packages/fcm-agent-core` / `packages/fcm-opencode` is still in progress (tracked in `.kandown/tasks/t2.md`). This release ships the working adapters and the extracted model-safety module ahead of that restructure.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "free-coding-models",
-  "version": "0.5.45",
+  "version": "0.5.47",
   "description": "Find the fastest coding LLM models in seconds — ping free models from multiple providers, pick the best one for OpenCode, Cursor, or any AI coding assistant.",
   "keywords": [
     "nvidia",

package/sources.js

@@ -404,20 +404,10 @@ export const kilo = [
 // 📖 Free unauthenticated tier works with tight shared limits; optional free token at https://token.llm7.io
 // 📖 Pro-tagged models from /v1/models are intentionally excluded.
 export const llm7 = [
-  // ── S+ tier — SWE-bench Verified ≥70% ──
-  ['deepseek-v4-flash', 'DeepSeek V4 Flash', 'S+', '79.0%', '1M'],
-  ['kimi-k2.6', 'Kimi K2.6', 'S+', '80.2%', '240k'],
-  ['claude-opus-4-8', 'Claude Opus 4.8', 'S+', '88.6%', '1M'],
-  ['claude-sonnet-5', 'Claude Sonnet 5', 'S+', '-', '1M'],
-  ['claude-fable-5', 'Claude Fable 5', 'S+', '95.0%', '1M'],
-  ['gpt-5.5', 'GPT-5.5', 'S+', '88.7%', '1.05M'],
-  ['gpt-5.4', 'GPT-5.4', 'S+', '80.0%', '1.05M'],
-  ['glm-5.2', 'GLM 5.2', 'S+', '82.8%', '1M'],
-  // ── A tier — SWE-bench Verified 40–50% ──
+  // 📖 LLM7 live /v1/models: only `turbo` tier is free (noKeyNeeded). All `pro` models are usage-based paid.
+  // ── S tier — SWE-bench Verified 60–70% ──
   ['devstral-small-2:24b', 'Devstral Small 2', 'S', '68.0%', '255k'],
-  ['gpt-5.4-mini', 'GPT-5.4 mini', 'A', '-', '400k'],
-  ['minimax-m2.7', 'MiniMax M2.7', 'S+', '78.0%', '180k'],
-  // ── B+ tier — SWE-bench Verified 30–35% ──
+  // ── A tier — SWE-bench Verified 40–50% ──
   ['codestral-latest', 'Codestral Latest', 'A', '40.0%', '32k'],
 ]