free-coding-models 0.3.67 → 0.3.68

package/changelog/v0.3.37.md

@@ -0,0 +1,182 @@
+# Changelog v0.3.37 - 2026-04-09
+
+### Changed
+- **Removed X/Twitter references** — no more `🐦 Follow me on X` links in README and TUI footer
+- **Discord link more prominent** in TUI footer (replaced X banner)
+- **README navigation bar** — now links all 10 sections (Why, Quick Start, Providers, Usage, TUI Keys, Features, Contributing, Licensing, Security, Support, License)
+- **Deleted accidental `echo` file** from repo
+- 238 models across 25 providers
+
+## [0.3.35] - 2026-04-07
+
+### Added
+- **OVHcloud AI Endpoints** — new European sovereign AI provider (8 models: Qwen3 Coder 30B MoE, GPT OSS 120B, GPT OSS 20B, Llama 3.3 70B, Qwen3 32B, R1 Distill 70B, Mistral Small 3.2, Llama 3.1 8B)
+- Free sandbox mode: 2 req/min per IP per model (no API key needed), 400 RPM with API key
+- **Now 238 models across 25 providers** (was 230/24)
+
+### SECURITY.md
+- **SECURITY.md** — full security policy with vulnerability reporting, architecture, and supply chain docs
+- **CODEOWNERS** — all changes require @vava-nessa review
+- **Dependabot** — weekly automated dependency + GitHub Actions updates (`.github/dependabot.yml`)
+- **Security Audit CI** — `npm audit` + lockfile lint on every push/PR + weekly schedule (`.github/workflows/security-audit.yml`)
+- **npm Provenance** — release workflow now publishes with `--provenance` (Sigstore-signed)
+- **SBOM generation** — Software Bill of Materials attached to every GitHub Release
+- **README trust badges** — dependency count, provenance, supply chain badges
+- **README 🛡️ Security section** — what the tool does/doesn't do, supply chain table
+
+## [0.3.34] - 2026-04-06
+
+### Added
+- **Chutes AI** — new decentralized free provider (4 models: DeepSeek R1, Llama 3.1 70B, Qwen 2.5 72B, Qwen2.5 Coder 32B)
+- **Google Gemini 3.1 Pro** — replaced dead Gemini 3 Pro (shut down March 9)
+- **Google Gemma 4 family** — Gemma 4 31B, Gemma 4 26B MoE, Gemma 4 E4B added to Google AI Studio
+- **Qwen3.6 Plus** and **Qwen3.5 Plus** added to Alibaba DashScope
+- **NVIDIA NIM: Kimi K2 Instruct 0905** added
+- **SambaNova: Qwen3-235B-A22B-Instruct-2507** added
+- **OpenRouter: ~10 new free models** (arcee trinity, mimo-v2-flash, etc.) — 25 total
+- **Together AI: expanded to 19 free models** (DeepSeek V3.2, MiniMax M2.1, etc.)
+- **Cloudflare: 4 new models** (gemma-4-26b, mistral-small-3.1, qwq-32b, granite-4.0) — 15 total
+- **Scaleway: gpt-oss-120b and holo2-30b-a3b** added
+- **Hyperbolic: gpt-oss-20b and Qwen3-235B-Instruct-2507** added
+- **Rovo: GPT-5.2, GPT-5.2-Codex, Claude Haiku 4.5** added
+- **OpenCode Zen: mimo-v2-pro-free, mimo-v2-omni-free** added; fixed context windows (gpt-5-nano: 128k→400k, nemotron-3-super-free: 128k→1M)
+
+### Changed
+- **Groq scout model renamed** — `llama-4-scout-17b-16e-preview` → `llama-4-scout-17b-16e-instruct`
+- **Now 230 models across 24 providers** (was 174/23)
+- **README provider table fully updated** with accurate per-provider counts
+
+### Fixed
+- Removed 3 deprecated Cerebras models (llama3.3-70b, qwen-3-32b, llama-4-scout-17b-16e-instruct) — 4 remain
+- Fixed missing comma in googleai array causing undefined model entry
+- Replaced all `??%` placeholder SWE scores with reasonable estimates
+
+## [0.3.33] - 2026-04-01
+
+### Changed
+- **X badge darker fuchsia background** (`rgb(140,0,80)`) for better readability
+- **Updated text** → "Follow me on X: @vavanessadev to check my other projects! 💖"
+
+## [0.3.32] - 2026-04-01
+
+### Fixed
+- **Corrected X/Twitter username to @vavanessadev** — link and display text now point to the right profile
+
+### Changed
+- **X support badge now on fuchsia background** — bold white-on-fuchsia with yellow username, impossible to miss
+- **Discord moved to the very bottom line** of the TUI footer
+- **X follow link added to top of README**
+
+## [0.3.31] - 2026-04-01
+
+### Changed
+- **Footer layout reorganized** — Discord link moved to the very bottom line of the TUI for better visibility
+- **X/Twitter support message** — added 🐦 "Support me on X: @vavanessadev 💖" with clickable link in the last release footer line
+
+## [0.3.29] - 2026-03-27
+
+### Added
+- **Fluorescent green UPDATE AVAILABLE banner** — impossible-to-miss fluo green (🚀⬆️) banner at the bottom of the TUI when a new version is detected; click it to update instantly
+- **Command Palette update entry** — `⬆️ UPDATE NOW` is always the first result in Ctrl+P when an update is available
+- **Shift+U hotkey** — press Shift+U from the main table to trigger an immediate update
+- **Mouse-clickable update banner** — click the fluo green banner to install the latest version and relaunch
+- **Background version re-check every 5 minutes** — if a new version is published while the TUI is open, the banner appears live without restarting
+- **Aggressive pre-TUI update prompt** — fluorescent green header, and "Continue without update" warns that reminders will follow
+- **Last release date in footer** — light pink `Last release: Mar 27, 2026, 09:42 PM` shows when the package was last published to npm, so users know how fresh the model data is
+- **Demo2 GIF** — added `demo2.gif` to README for visual showcase of the TUI
+
+### Fixed
+- **Auto-update now detects the correct package manager** — bun, pnpm, and yarn users no longer get duplicate npm installs (fixes #46)
+- Update banner in footer shows the correct install command for your package manager
+
+## [0.3.28] - 2026-03-27
+
+_(Skipped — published as 0.3.28 without changelog in commit message, all changes are in 0.3.29)_
+
+## [0.3.27] - 2026-03-27
+
+## [0.3.26] - 2026-03-27
+
+### Added
+- **Groq**: Added Compound + Compound Mini; fixed Llama 4 Scout context (10M)
+- **OpenRouter**: Added MiniMax M2.5, Nemotron 3 Super, Hermes 3 405B, Gemma 3n E4B
+- **HuggingFace**: Replaced invalid DeepSeek-V3-Coder + outdated starcoder2-15b with DeepSeek V3 0324 + Qwen2.5 Coder 32B
+- **Replicate**: Replaced CodeLlama 70B (2023) with DeepSeek V3 0324 + Llama 3.3 70B
+- **Cloudflare**: Added Kimi K2.5, GLM-4.7-Flash, Llama 4 Scout, Nemotron 3 Super, Qwen3 30B MoE
+- **Scaleway**: Added Qwen3.5 400B VLM + Mistral Large 675B
+- **DeepInfra**: Replaced Mixtral Code with Nemotron 3 Super + DeepSeek V3 0324 + Qwen3 235B
+- **Fireworks**: Added Llama 4 Maverick + Qwen3 235B
+- **Hyperbolic**: Added Qwen3 80B Thinking variant
+- **Together AI**: Added Qwen3.5 400B VLM, MiniMax M2.5, GLM-5
+
+### Changed
+- **Rovo Dev CLI**: Updated Sonnet 4 → Sonnet 4.6, added Opus 4.6
+- **Groq**: Removed 4 deprecated models (R1 Distill 70B, QwQ 32B, Kimi K2, Maverick)
+- **OpenRouter**: Updated context sizes for multiple models
+
+## [0.3.25] - 2026-03-19
+
+### Added
+- **Installed Models Manager** — View, launch, and disable models configured in external tools (Goose, Crush, Aider, Qwen, Pi, OpenHands, Amp)
+  - Access via Command Palette (Ctrl+P) → "Installed models"
+  - Scans all supported tool configs automatically on opening
+  - Displays all models per tool (e.g., Crush shows both large and small models)
+  - Actions: Launch (Enter), Disable (D) with backup, Reinstall (R)
+  - Soft delete: Comments out model entries and saves backups to ~/.free-coding-models-backups.json
+- **Full mouse support for the TUI** — Click, right-click, double-click, and scroll work throughout the interface
+  - **Click column headers** — Sort by any column (click Rank, Tier, SWE%, CTX, Model, Provider, etc.)
+  - **Click model rows** — Move cursor to any model (left-click)
+  - **Right-click model rows** — Toggle favorite (same as F key)
+  - **Double-click model rows** — Select model and launch (same as Enter)
+  - **Mouse wheel** — Scroll through the main table, overlays (Settings, Help, Changelog), and command palette results
+  - **Click CLI Tools header** — Cycle through tool modes (same as Z key)
+  - **Click Tier header** — Cycle through tier filters (same as T key)
+  - **Click footer hotkeys** — Trigger any visible hotkey from the footer
+  - **Command palette click** — Click inside to select items, double-click to confirm; click outside to close
+  - **Recommend questionnaire click** — Click on option rows to select, double-click to confirm
+- **Mouse unit tests** — 46 new tests covering SGR sequence parsing, double-click detection, modifiers, and COLUMN_SORT_MAP validation
+
+### Changed
+- **CLI Tools column redesigned** — Renamed from "Compatible with" to "CLI Tools", with left-aligned emoji display (compatible tools packed left instead of fixed slot positions)
+- **Sort arrow overflow fixed** — SWE%, CTX, Stability, and Uptime columns now properly fit within their widths when sorted (arrow now `↑SWE%` instead of `↑ SWE%`)
+- **Mouse sequence suppression** — SGR mouse sequences no longer leak into keypress handlers (prevents spurious sort/filter triggers when clicking)
+
+### Fixed
+- **Command palette scroll leak** — Mouse wheel no longer injects raw SGR sequence bytes into the command palette text input
+- **Double-action on model click** — Clicking a model row now only moves the cursor; it no longer simultaneously triggers column sorting
+- **Mouse event listener order** — Fixed race condition where readline emitted keypress events before mouse data was processed
+
+## [0.3.24] - 2026-03-19
+
+### Added
+- **Unique emoji per tool** — Every CLI tool now has a dedicated emoji shown in the Compatible column, Z-cycle badge, command palette, help overlay, and README (📦 OpenCode, 🦞 OpenClaw, 💘 Crush, 🪿 Goose, π Pi, 🛠 Aider, 🐉 Qwen, 🤲 OpenHands, ⚡ Amp, 🦘 Rovo, ♊ Gemini)
+- **Merged compat column** — OpenCode CLI and Desktop share 📦 in a single slot (11 slots instead of 12 separate initials)
+- **COMPAT_COLUMN_SLOTS** — New export in tool-metadata.js for merged compatible-column rendering
+- **Width warning now always shows** - Terminal width warning displays every time terminal is resized below 80 columns (previously limited to 2 shows per session)
+- **Gemini CLI integration** - New CLI-only tool provider with 3 models (Gemini 3 Pro 🆕, Gemini 2.5 Pro, Gemini 2.5 Flash)
+- **Rovo Dev CLI integration** - New CLI-only tool provider with Claude Sonnet 4 🆕
+- **Tool compatibility alerts** - When trying to launch Rovo/Gemini models with wrong tool, shows alert and offers to switch
+- **Auto-install detection** - Prompt to install CLI tools when binary not found (Rovo/Gemini)
+- **OpenAI-compatible API support for Gemini** - Gemini CLI can use custom providers via environment variables
+- **New CLI flags** - Added `--rovo` and `--gemini` launch options
+- **"🆕" badges** - Mark newly added models in the table (Claude Sonnet 4, Gemini 3 Pro)
+- **OpenCode Zen free models** - 5 new free models (Big Pickle, GPT 5 Nano, MiMo V2 Flash Free, MiniMax M2.5 Free, Nemotron 3 Super Free) exclusive to OpenCode CLI/Desktop via `opencode-zen` provider
+- **"Compatible with" column** - New TUI column showing colored emojis for each tool a model supports; incompatible tools show dim spaces
+- **Tool color system** - Each of the 12 supported tools now has a unique RGB color and emoji used in the Z-cycle badge and compatibility column
+- **Incompatible model highlighting** - When a tool mode is active (via Z), models that can't run with that tool get a dark red background for instant visibility — they stay in their normal sorted position (not pushed to the bottom)
+- **Tool compatibility functions** - `getCompatibleTools()` and `isModelCompatibleWithTool()` in tool-metadata.js for programmatic compatibility checks
+- **Incompatible model fallback overlay** - When pressing Enter on a model that can't run on the active tool (red-highlighted row), an in-TUI overlay appears with two options: (1) switch to a compatible tool, or (2) pick a similar model by SWE score that works with the current tool
+- **findSimilarCompatibleModels()** - New function in tool-metadata.js that finds models with closest SWE scores compatible with the active tool
+- **Updated provider/model counts** - Now 23 providers with 171 models (was 20/160)
+
+### Changed
+- **Z key cycle** - Rovo and Gemini added to tool mode cycle (last in order)
+- **Tool metadata** - Removed `initial` field (replaced by emojis), added `cliOnly` flag for CLI-only tools, `emoji` and `color` properties for all 12 tools
+- **Provider metadata** - Added Rovo, Gemini, and OpenCode Zen provider information
+- **Responsive column hiding** - Compatible column hides first (before Rank) on narrow terminals
+- **Key handler** - Zen models auto-switch to OpenCode CLI on launch; API key warnings skip Zen models
+- **Documentation** - Updated README with CLI-only tools section, Zen models, compatibility matrix
+
+### Fixed
+- **Missing import error** - Fixed `getToolMeta` not defined in key-handler.js
+- **CLI-only tools API key requirement** - Gemini CLI and Rovo Dev CLI no longer require API keys to be configured before launching; these tools manage their own authentication

package/changelog/v0.3.38.md

@@ -0,0 +1,3 @@
+# Changelog v0.3.38 - 2026-04-09
+
+Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

package/changelog/v0.3.39.md

@@ -0,0 +1,3 @@
+# Changelog v0.3.39 - 2026-04-09
+
+0.3.39 - 🔧 Fix CI/CD: switch GitHub Actions from npm ci to pnpm

package/changelog/v0.3.4.md

@@ -0,0 +1,9 @@
+# Changelog v0.3.4 - 2026-03-16
+
+### Added
+- **Proxy root landing JSON**: `GET /` on FCM Proxy V2 now returns a small unauthenticated status payload, so browser checks no longer fail with `{"error":"Unauthorized"}`.
+- **`daemon stop` CLI command**: The public CLI now supports `free-coding-models daemon stop`, matching the existing daemon manager capability and the documented workflow.
+
+### Fixed
+- **README/UI parity restored**: The docs now match the current product surface, including `160` models, the `Used` token-history column, and the current launcher/proxy behavior.
+- **Malformed config sections are normalized on load**: Invalid `apiKeys`, `providers`, or `settings` values are now coerced back to safe empty objects instead of leaking broken runtime shapes into the app.

package/changelog/v0.3.40.md

@@ -0,0 +1,18 @@
+# Changelog v0.3.40 - 2026-04-09
+
+### Fixed
+- **🔧 CI/CD** — Added `packageManager` field to `package.json` so `pnpm/action-setup@v4` can resolve the pnpm version
+
+## [0.3.39] - 2026-04-09
+
+### Fixed
+- **🔧 CI/CD** — Switched GitHub Actions from `npm ci` to `pnpm install --frozen-lockfile` (release + security-audit workflows)
+- Removed obsolete `package-lock.json` that was causing `EUSAGE` sync errors in CI
+
+## [0.3.38] - 2026-04-09
+
+### Added
+- **🔮 Hermes Agent** — Configures Hermes via `hermes config set`, restarts the gateway, and launches `hermes chat`. Supports all OpenAI-compatible providers.
+- **▶️ Continue CLI** — Writes `~/.continue/config.yaml` with `provider: openai` + `apiBase` and launches `cn`.
+- **🧠 Cline CLI** — Writes `~/.cline/globalState.json` with OpenAI-compatible config and launches `cline`.
+- `--hermes`, `--continue`, `--cline` CLI flags, `Z` cycle entries, command palette actions, and Install Endpoints support for all three tools.

package/changelog/v0.3.41.md

@@ -0,0 +1,7 @@
+# Changelog v0.3.41 - 2026-04-09
+
+### Added
+- Added support for Xcode Intelligence (Xcode 26.3+) via the `--xcode` flag and manual configuration flow. The tool provides clear setup instructions and launches Xcode directly.
+- The `xcode` launcher mode has been registered in the tool metadata, enabling Xcode selection through the command palette, `Z` cycle, and CLI arguments.
+- Users can now use Xcode Intelligence with any of the 230+ supported OpenAI-compatible API endpoints.
+- Compatibility logic ensures Xcode mode accurately highlights supported models (compatible with all standard models).

package/changelog/v0.3.42.md

@@ -0,0 +1,9 @@
+# Changelog v0.3.42 - 2026-04-09
+
+### Added
+- Added a Star History section to the README so users can quickly see project growth and adoption over time.
+
+### Fixed
+- Xcode Intelligence bootstrap detection now recognizes `xcodebuild`, so the launcher can detect a valid local Xcode install instead of treating the tool as missing.
+- `free-coding-models --web` now checks whether port `3333` already belongs to another app and automatically starts on the next free local port instead of sending users to a misleading `Not Found` page.
+- npm releases now build the web dashboard during `prepack`, so published installs include `web/dist` and the web UI actually loads after install instead of starting a server with no bundled frontend.

package/changelog/v0.3.43.md

@@ -0,0 +1,13 @@
+# Changelog v0.3.43 - 2026-04-09
+
+### Added
+- Added new anonymous telemetry events for real product usage: `app_use` records which tool, provider, and model users actually launch, while `app_action` tracks a small set of high-signal in-app actions such as endpoint installs, API key save/remove flows, and shell env export toggles.
+- Added per-session telemetry linking with a generated `session_id`, so `app_start`, `app_use`, and action events can be analyzed together while still keeping users anonymous through the existing persistent install-scoped `distinct_id`.
+- Added launch outcome tracking with `app_use_result`, so the dashboard can distinguish successful launches from blocked flows like incompatible model/tool combinations or missing CLI installs.
+
+### Changed
+- Extended the telemetry pipeline to support custom event properties on top of the shared anonymous metadata (`app`, `version`, `mode`, `system`, `terminal`) without introducing a second analytics path.
+- Updated the README telemetry section to explain that anonymous analytics now cover launched tool/provider/model metadata and a few product actions, while still explicitly excluding prompts, source code, file paths, API keys, and secrets.
+
+### Fixed
+- Added regression coverage for telemetry opt-out behavior and event payload shape, ensuring the new launch/action events stay disabled under `--no-telemetry` or `FREE_CODING_MODELS_TELEMETRY=0`.

package/changelog/v0.3.44.md

@@ -0,0 +1,10 @@
+# Changelog v0.3.44 - 2026-04-10
+
+### Added
+- Added **jcode** external tool support, integrating a new CLI tool across the codebase with command-palette description, tool metadata (label, emoji, flag, color), bootstrap metadata (binary, docs URL, install commands), and full launch wiring via `prepareExternalToolLaunch` and `startExternalTool`. CLI flag parsing for `--jcode` (jcodeMode) is also included.
+
+### Changed
+- Provider API keys are now synced into tool launches — `openclaw` imports `getApiKey` and `syncShellEnv`, accepts an env override in `spawnOpenClawCli`, and populates the child process env with the provider API key when launching the CLI. Tool launchers now use `model.providerKey` instead of hardcoded provider IDs, and `prepareExternalToolLaunch` includes provider and API key in launch args. This enables multi-provider support and ensures launched tools can authenticate using the configured key.
+
+### Fixed
+- README now includes a "Bonus Free Stuff" section with curated resources: community awesome-lists, AI-powered IDEs with free tiers, API providers with permanent free tiers, trial credit providers, and education/developer program freebies — accessible via a new navigation link.

package/changelog/v0.3.45.md

@@ -0,0 +1,12 @@
+# Changelog v0.3.45 - 2026-04-11
+
+### Added
+- Added **jcode** external tool support, integrating a new CLI tool across the codebase with command-palette description, tool metadata (label, emoji, flag, color), bootstrap metadata (binary, docs URL, install commands), and full launch wiring via `prepareExternalToolLaunch` and `startExternalTool`. CLI flag parsing for `--jcode` (jcodeMode) is also included.
+- Added **Alt+W** footer toggle — collapses the footer to a single-line hint showing only the toggle hotkey and exit command, giving more vertical space for the model table.
+
+### Changed
+- Provider API keys are now synced into tool launches — `openclaw` imports `getApiKey` and `syncShellEnv`, accepts an env override in `spawnOpenClawCli`, and populates the child process env with the provider API key when launching the CLI. Tool launchers now use `model.providerKey` instead of hardcoded provider IDs, and `prepareExternalToolLaunch` includes provider and API key in launch args. This enables multi-provider support and ensures launched tools can authenticate using the configured key.
+- Command Palette label updated to remove "NEW !" badge — the feature is no longer new.
+
+### Fixed
+- README now includes a "Bonus Free Stuff" section with curated resources: community awesome-lists, AI-powered IDEs with free tiers, API providers with permanent free tiers, trial credit providers, and education/developer program freebies — accessible via a new navigation link.

package/changelog/v0.3.46.md

@@ -0,0 +1,5 @@
+# Changelog v0.3.46 - 2026-04-11
+
+### Changed
+- **jcode** moved to 3rd position in the `Z` cycle order (was 17th/last) — now cycles as: OpenCode → Pi → jcode → OpenCode Desktop → OpenClaw → Crush → Goose → Aider → Qwen → OpenHands → Amp → Hermes → Continue → Cline → Xcode → Rovo → Gemini
+- **jcode launch wiring overhaul** — improved provider handling: NVIDIA now uses `OPENAI_API_KEY` env var instead of `OPENAI_API_BASE`, and all other providers now use `--provider openai-compatible` with explicit `OPENAI_BASE_URL` + `OPENAI_API_KEY` in the env, fixing multi-provider authentication for jcode launches

package/changelog/v0.3.47.md

@@ -0,0 +1,4 @@
+# Changelog v0.3.47 - 2026-04-11
+
+### Fixed
+- **jcode model validation freeze** — Fixed terminal freeze when launching jcode with custom models (e.g. GPT OSS 120B via NVIDIA NIM). jcode has a hardcoded model whitelist that rejected non-standard model names passed via `--model` flag, causing the REPL to start without a valid model and hang unresponsively. Now uses `JCODE_MODEL` env var instead, which bypasses the validation and allows any model name to work correctly.

package/changelog/v0.3.48.md

@@ -0,0 +1,10 @@
+# Changelog v0.3.48 - 2026-04-11
+
+### Fixed
+- **jcode model validation freeze (for real this time)** — jcode has a hardcoded model whitelist that rejects bare model names like `gpt-oss-120b` via `--model` flag. The v0.3.47 fix using `JCODE_MODEL` env var didn't actually work — jcode silently ignores it for the `openai-compatible` provider. The real fix uses two strategies:
+  1. **Native providers**: For providers that jcode supports natively (Groq, Cerebras, DeepInfra, Scaleway, Together, Hugging Face, Fireworks, Chutes, OpenRouter, Perplexity, ZAI, Mistral), we now use `--provider groq` instead of `--provider openai-compatible`. Native providers accept namespaced model names like `openai/gpt-oss-120b` without validation issues.
+  2. **OpenAI-compatible fallback**: For providers without a native jcode match (NVIDIA NIM, SambaNova, etc.), we keep `--provider openai-compatible` but ensure model IDs always have a namespace prefix (`openai/gpt-oss-120b` instead of `gpt-oss-120b`) and set a placeholder `OPENROUTER_API_KEY` to satisfy jcode's false credential check on namespaced models.
+
+### Added
+- **`JCODE_NATIVE_PROVIDERS` mapping** — New provider mapping table that routes 12 of our providers to jcode's native provider system, with correct env var names extracted from the jcode binary. This gives better compatibility and avoids the `openai-compatible` provider's model validation bugs.
+- **`ensureJcodeModelPrefix()` helper** — Ensures model IDs always have a namespace prefix (e.g. `gpt-oss-120b` → `openai/gpt-oss-120b`) so jcode's whitelist validation is bypassed.

package/changelog/v0.3.49.md

@@ -0,0 +1,18 @@
+# Changelog v0.3.49 - 2026-04-11
+
+### Changed
+
+- **API key validation now uses provider auth endpoints instead of pings** — Previously, testing an API key sent a chat completion `POST` to a model endpoint and waited for a 200 response. This was indirect (a model being down didn't mean the key was invalid) and slow (sequential probes with 4s delays).
+
+- **Parallel auth-only probes (3×8s)** — For providers that expose an OpenAI-compatible `/v1/models` or a `/v1/account` endpoint, the test now fires 3 simultaneous probes to that endpoint. A 200 response = key valid and accepted. A 401/403 response = key rejected (auth error). This is decisive and ~8× faster than the ping approach.
+
+- **Parallel ping burst fallback** — Providers without an auth-check endpoint (replicate, cloudflare, zai, googleai, opencode-zen, rovo, gemini) or that timed out now fall back to chat completion pings, but in **parallel batches of 5 probes** instead of 1-by-1 with 4s delays. This keeps the overall test under 30s even for "difficult" providers.
+
+- **Auto-test all keys on Settings open (`P`)** — Opening Settings now automatically fires parallel auth probes for all configured API keys simultaneously. Users see the status of every provider the moment they open the screen, instead of having to press `T` for each one individually.
+
+### Providers with dedicated auth endpoints
+
+| Provider | Auth check endpoint | Method |
+|----------|-------------------|--------|
+| NVIDIA NIM | `/v1/account` | GET |
+| Groq, Cerebras, SambaNova, OpenRouter, HuggingFace, DeepInfra, Fireworks, Hyperbolic, Scaleway, SiliconFlow, Together, Perplexity, Chutes, OVHcloud, Qwen, iFlow | `/v1/models` | GET |

package/changelog/v0.3.5.md

@@ -0,0 +1,5 @@
+# Changelog v0.3.5 - 2026-03-16
+
+### Fixed
+- **Claude Code beta-route compatibility**: FCM Proxy V2 now matches routes on the URL pathname, so Anthropic requests like `/v1/messages?beta=true` and `/v1/messages/count_tokens?beta=true` resolve correctly instead of failing with a fake “selected model may not exist” error.
+- **Claude proxy parity with `free-claude-code`**: The Claude integration was revalidated against the real `claude` binary, and the proxy-side Claude alias mapping now reaches the upstream provider again in the exact `free-claude-code` style flow.

package/changelog/v0.3.50.md

@@ -0,0 +1,41 @@
+# Changelog v0.3.50 - 2026-04-11
+
+### Changed
+
+- **Providers reordered by generosity of free tier** — All 25 providers are now sorted from most generous to least generous in the README, TUI Settings page, and `D` key filter cycling. No more hunting for the best free option.
+
+- **Free tier limits corrected across all providers** — Verified and corrected free tier limits for every provider using live web research. Key corrections:
+  - **Groq**: 30 RPM, 1K-14.4K req/day (previously listed as "30-50 RPM per model")
+  - **Google AI Studio**: 15-60 RPM, 250-1.5K req/day (previously listed as "14.4K req/day, 30/min")
+  - **Together AI**: ❌ **No free tier** — requires $5 minimum purchase. Removed from the "free" recommendation.
+  - **iFlow**: ⚠️ **Shutting down April 17, 2026** — marked with warning in README and sources.js
+
+- **README subtitle updated** — Now says "ranked by generosity of free tier (most generous first)" instead of "ranked by SWE-bench"
+
+### Provider generosity ranking (most generous first)
+
+1. Groq (30 RPM, 1K-14.4K req/day)
+2. Cerebras (1M tokens/day)
+3. Google AI Studio (15-60 RPM, 250-1.5K req/day)
+4. NVIDIA NIM (~40 RPM)
+5. Cloudflare Workers AI (10K neurons/day)
+6. OpenRouter (50 req/day free)
+7. DeepInfra (200 concurrent requests)
+8. HuggingFace (~$0.10/month)
+9. Perplexity (~50 RPM tiered)
+10. SambaNova (generous dev quota)
+11. Fireworks AI ($1 credits)
+12. Hyperbolic ($1 credits)
+13. OVHcloud AI (2 req/min/IP free)
+14. Replicate (6 req/min free)
+15. Codestral (30 RPM, 2K req/day)
+16. ZAI (generous free quota)
+17. Scaleway (1M tokens)
+18. Alibaba DashScope (1M tokens/90 days)
+19. SiliconFlow (100 req/day + $1 credits)
+20. Rovo Dev CLI (5M tokens/day)
+21. Gemini CLI (1K req/day)
+22. Chutes AI (free community GPU)
+23. OpenCode Zen (free with account)
+24. Together AI (❌ no free tier)
+25. iFlow (⚠️ shutting down April 17, 2026)

package/changelog/v0.3.51.md

@@ -0,0 +1,37 @@
+# Changelog v0.3.51 - 2026-04-11
+
+### Changed
+
+- **NVIDIA NIM moved to #1** — Now listed first in README, TUI Settings page, and `D` key filter cycling (per user request). Provider order across all surfaces is now: NVIDIA NIM → Groq → Cerebras → Google AI Studio → Cloudflare → ... → iFlow.
+
+### Added
+
+- **Provider generosity ranking** — README subtitle now includes "ranked by free tier generosity" and the full 25-provider ranking table. This reflects the same order used in the TUI Settings screen and `D` key cycling.
+
+### Provider order (as shown in TUI and README)
+
+1. NVIDIA NIM (~40 RPM, 46 models)
+2. Groq (30 RPM, 1K-14.4K req/day, 8 models)
+3. Cerebras (30 RPM, 1M tokens/day, 4 models)
+4. Google AI Studio (15-60 RPM, 250-1.5K req/day, 6 models)
+5. Cloudflare Workers AI (10K neurons/day, 15 models)
+6. OpenRouter (50 req/day free, 25 models)
+7. DeepInfra (200 concurrent requests, 4 models)
+8. HuggingFace (~$0.10/month, 2 models)
+9. Perplexity (~50 RPM tiered, 4 models)
+10. SambaNova (generous dev quota, 13 models)
+11. Fireworks AI ($1 credits, 4 models)
+12. Hyperbolic ($1 credits, 13 models)
+13. OVHcloud AI (2 req/min/IP free, 8 models)
+14. Replicate (6 req/min free, 2 models)
+15. Codestral (30 RPM, 2K req/day, 1 model)
+16. ZAI (generous free quota, 7 models)
+17. Scaleway (1M tokens, 10 models)
+18. Alibaba DashScope (1M tokens/90 days, 11 models)
+19. SiliconFlow (100 req/day + $1 credits, 6 models)
+20. Rovo Dev CLI (5M tokens/day, 5 models)
+21. Gemini CLI (1K req/day, 3 models)
+22. Chutes AI (free community GPU, 4 models)
+23. OpenCode Zen (free with account, 7 models)
+24. Together AI (❌ no free tier, 19 models)
+25. iFlow (⚠️ shutting down April 17, 2026, 11 models)

package/changelog/v0.3.52.md

@@ -0,0 +1,43 @@
+# Changelog v0.3.52 - 2026-04-18
+
+### Added
+
+- **OpenCode WebUI Support** — Added `--opencode-web` flag to open the OpenCode WebUI dashboard after configuring the selected model. This mirrors the existing `--opencode-desktop` behavior.
+
+## [0.3.51] - 2026-04-11
+
+### Changed
+
+- **NVIDIA NIM moved to #1** — Now listed first in README, TUI Settings page, and `D` key filter cycling (per user request). Provider order across all surfaces is now: NVIDIA NIM → Groq → Cerebras → Google AI Studio → Cloudflare → ... → iFlow.
+
+### Added
+
+- **Provider generosity ranking** — README subtitle now includes "ranked by free tier generosity" and the full 25-provider ranking table. This reflects the same order used in the TUI Settings screen and `D` key cycling.
+
+### Provider order (as shown in TUI and README)
+
+1. NVIDIA NIM (~40 RPM, 46 models)
+2. Groq (30 RPM, 1K-14.4K req/day, 8 models)
+3. Cerebras (30 RPM, 1M tokens/day, 4 models)
+4. Google AI Studio (15-60 RPM, 250-1.5K req/day, 6 models)
+5. Cloudflare Workers AI (10K neurons/day, 15 models)
+6. OpenRouter (50 req/day free, 25 models)
+7. DeepInfra (200 concurrent requests, 4 models)
+8. HuggingFace (~$0.10/month, 2 models)
+9. Perplexity (~50 RPM tiered, 4 models)
+10. SambaNova (generous dev quota, 13 models)
+11. Fireworks AI ($1 credits, 4 models)
+12. Hyperbolic ($1 credits, 13 models)
+13. OVHcloud AI (2 req/min/IP free, 8 models)
+14. Replicate (6 req/min free, 2 models)
+15. Codestral (30 RPM, 2K req/day, 1 model)
+16. ZAI (generous free quota, 7 models)
+17. Scaleway (1M tokens, 10 models)
+18. Alibaba DashScope (1M tokens/90 days, 11 models)
+19. SiliconFlow (100 req/day + $1 credits, 6 models)
+20. Rovo Dev CLI (5M tokens/day, 5 models)
+21. Gemini CLI (1K req/day, 3 models)
+22. Chutes AI (free community GPU, 4 models)
+23. OpenCode Zen (free with account, 7 models)
+24. Together AI (❌ no free tier, 19 models)
+25. iFlow (⚠️ shutting down April 17, 2026, 11 models)

package/changelog/v0.3.54.md

@@ -0,0 +1,55 @@
+# Changelog v0.3.54 - 2026-04-18
+
+### Changed
+
+- **Improved OpenCode WebUI Launch** — The `--opencode-web` flag now correctly spawns the `opencode web` command, providing a integrated browser-based coding experience with pre-configured model selection.
+
+### Added
+
+- **Kilo CLI Support** — Added `--kilo` flag to launch the Kilo CLI with the selected model. Kilo is a fork of OpenCode and shares the same configuration structure (stored in `~/.config/kilo/opencode.json`).
+
+## [0.3.53] - 2026-04-18
+
+### Added
+
+- **OpenCode WebUI Support** — Added `--opencode-web` flag to open the OpenCode WebUI dashboard after configuring the selected model.
+
+## [0.3.52] - 2026-04-18
+
+## [0.3.51] - 2026-04-11
+
+### Changed
+
+- **NVIDIA NIM moved to #1** — Now listed first in README, TUI Settings page, and `D` key filter cycling (per user request). Provider order across all surfaces is now: NVIDIA NIM → Groq → Cerebras → Google AI Studio → Cloudflare → ... → iFlow.
+
+### Added
+
+- **Provider generosity ranking** — README subtitle now includes "ranked by free tier generosity" and the full 25-provider ranking table. This reflects the same order used in the TUI Settings screen and `D` key cycling.
+
+### Provider order (as shown in TUI and README)
+
+1. NVIDIA NIM (~40 RPM, 46 models)
+2. Groq (30 RPM, 1K-14.4K req/day, 8 models)
+3. Cerebras (30 RPM, 1M tokens/day, 4 models)
+4. Google AI Studio (15-60 RPM, 250-1.5K req/day, 6 models)
+5. Cloudflare Workers AI (10K neurons/day, 15 models)
+6. OpenRouter (50 req/day free, 25 models)
+7. DeepInfra (200 concurrent requests, 4 models)
+8. HuggingFace (~$0.10/month, 2 models)
+9. Perplexity (~50 RPM tiered, 4 models)
+10. SambaNova (generous dev quota, 13 models)
+11. Fireworks AI ($1 credits, 4 models)
+12. Hyperbolic ($1 credits, 13 models)
+13. OVHcloud AI (2 req/min/IP free, 8 models)
+14. Replicate (6 req/min free, 2 models)
+15. Codestral (30 RPM, 2K req/day, 1 model)
+16. ZAI (generous free quota, 7 models)
+17. Scaleway (1M tokens, 10 models)
+18. Alibaba DashScope (1M tokens/90 days, 11 models)
+19. SiliconFlow (100 req/day + $1 credits, 6 models)
+20. Rovo Dev CLI (5M tokens/day, 5 models)
+21. Gemini CLI (1K req/day, 3 models)
+22. Chutes AI (free community GPU, 4 models)
+23. OpenCode Zen (free with account, 7 models)
+24. Together AI (❌ no free tier, 19 models)
+25. iFlow (⚠️ shutting down April 17, 2026, 11 models)

package/changelog/v0.3.55.md

@@ -0,0 +1,62 @@
+# Changelog v0.3.55 - 2026-04-18
+
+### Changed
+
+- **Direct Launch Attempt** — Tool launchers for OpenCode and Kilo now attempt to spawn the command directly instead of blocking with a pre-installation check. Installation instructions are now provided via the process error handler for a more seamless experience.
+- **Fixed OpenCode NPM Command** — Updated the suggested installation command for OpenCode to `npm install -g opencode-ai`.
+
+## [0.3.54] - 2026-04-18
+
+### Changed
+
+- **Improved OpenCode WebUI Launch** — The `--opencode-web` flag now correctly spawns the `opencode web` command, providing a integrated browser-based coding experience with pre-configured model selection.
+
+### Added
+
+- **Kilo CLI Support** — Added `--kilo` flag to launch the Kilo CLI with the selected model. Kilo is a fork of OpenCode and shares the same configuration structure (stored in `~/.config/kilo/opencode.json`).
+
+## [0.3.53] - 2026-04-18
+
+### Added
+
+- **OpenCode WebUI Support** — Added `--opencode-web` flag to open the OpenCode WebUI dashboard after configuring the selected model.
+
+## [0.3.52] - 2026-04-18
+
+## [0.3.51] - 2026-04-11
+
+### Changed
+
+- **NVIDIA NIM moved to #1** — Now listed first in README, TUI Settings page, and `D` key filter cycling (per user request). Provider order across all surfaces is now: NVIDIA NIM → Groq → Cerebras → Google AI Studio → Cloudflare → ... → iFlow.
+
+### Added
+
+- **Provider generosity ranking** — README subtitle now includes "ranked by free tier generosity" and the full 25-provider ranking table. This reflects the same order used in the TUI Settings screen and `D` key cycling.
+
+### Provider order (as shown in TUI and README)
+
+1. NVIDIA NIM (~40 RPM, 46 models)
+2. Groq (30 RPM, 1K-14.4K req/day, 8 models)
+3. Cerebras (30 RPM, 1M tokens/day, 4 models)
+4. Google AI Studio (15-60 RPM, 250-1.5K req/day, 6 models)
+5. Cloudflare Workers AI (10K neurons/day, 15 models)
+6. OpenRouter (50 req/day free, 25 models)
+7. DeepInfra (200 concurrent requests, 4 models)
+8. HuggingFace (~$0.10/month, 2 models)
+9. Perplexity (~50 RPM tiered, 4 models)
+10. SambaNova (generous dev quota, 13 models)
+11. Fireworks AI ($1 credits, 4 models)
+12. Hyperbolic ($1 credits, 13 models)
+13. OVHcloud AI (2 req/min/IP free, 8 models)
+14. Replicate (6 req/min free, 2 models)
+15. Codestral (30 RPM, 2K req/day, 1 model)
+16. ZAI (generous free quota, 7 models)
+17. Scaleway (1M tokens, 10 models)
+18. Alibaba DashScope (1M tokens/90 days, 11 models)
+19. SiliconFlow (100 req/day + $1 credits, 6 models)
+20. Rovo Dev CLI (5M tokens/day, 5 models)
+21. Gemini CLI (1K req/day, 3 models)
+22. Chutes AI (free community GPU, 4 models)
+23. OpenCode Zen (free with account, 7 models)
+24. Together AI (❌ no free tier, 19 models)
+25. iFlow (⚠️ shutting down April 17, 2026, 11 models)