feat: dynamic model registry + plan-aware grouping, release 1.3.0

Feat: dynamic models
- Fetch live /provider/v1/models on refresh and merge with bundled list
- Serve dynamic /v1/models endpoint (new models appear without a release)
- Expanded registry 44 -> 52 models (Kimi-K3, Qwen 3.7 Flash, Gemini
  3.6 Flash/3.5 Flash Lite, Inkling, Inkling Small, Laguna, Ling)

Feat: plan-aware UI
- Page 5 groups by plan access first, then Free/Billing, then provider
- PlanAccess mirrors CLI evaluateModelAccess (go = open source + Luna/Grok,
  pro blocks Fable/Opus + Fugu, credits override)
- Page 7 groups by plan (Available on Go vs Requires Pro/Max)
- Cost sync result shows as status-bar notification

Fix: keymap
- o always copies OpenAI endpoint; clear-old moved to Shift+O

Chore: release
- Bump to 1.3.0 (package.json, pubspec.yaml, version.dart)
- Sync CHANGELOG + docs

Author: Khip01

AGENTS.md

@@ -31,13 +31,13 @@ commandcode-bridge/
 │       ├── main.dart                 # CLI wiring (run, run --server, cost-sync, help)
 │       ├── models/
 │       │   ├── account.dart          # Account + config store (port persist)
-│       │   ├── models_db.dart        # 44 models with goAccessible field
+│       │   ├── models_db.dart        # 52 bundled models + PlanAccess rules
 │       │   └── version.dart          # Bridge version constant
 │       ├── services/
-│       │   ├── api_client.dart       # HTTP client for api.commandcode.ai
+│       │   ├── api_client.dart       # HTTP client for api.commandcode.ai (incl. /provider/v1/models)
 │       │   ├── cost_sync.dart        # Cost sync: detect agents, update configs
 │       │   ├── log_store.dart        # JSONL activity log (2000 entries)
-│       │   ├── pricing_db.dart       # Hardcoded pricing table (44 models)
+│       │   ├── pricing_db.dart       # Hardcoded pricing table (52 models)
 │       │   └── updater.dart          # Self-update: API cache + download .tgz + npm install -g
 │       ├── server/
 │       │   ├── server_controller.dart # HTTP server + routing
@@ -167,7 +167,7 @@ npm v11 has a bug installing global git deps: the install appears to succeed (`a
 | `/v1/chat/completions` | POST | OpenAI-compatible chat completions |
 | `/v1/messages` | POST | Anthropic-compatible Messages API |
 | `/messages` | POST | Anthropic-compatible Messages API (alt path) |
-| `/v1/models` | GET | List 44 available models |
+| `/v1/models` | GET | List available models (live API merged with bundled, 52+) |
 | `/v1/health` | GET | Health check |
 | `/v1/token` | GET | Get access token |
 | `/v1/info` | GET | Bridge info + config |
@@ -217,18 +217,34 @@ npm v11 has a bug installing global git deps: the install appears to succeed (`a
 
 ## Plan Access
 
-Plan model access rules (from cli.mjs):
+Plan model access rules mirror the official Command Code CLI
+(`evaluateModelAccess` in `cli.mjs`). The bridge implements them in
+`PlanAccess` in `models_db.dart`:
 
-| Plan | Allowed Categories | Blocked Premium Models |
-|------|-------------------|----------------------|
-| `individual-go` | opensource only | (all premium blocked) |
-| `individual-pro` | premium + opensource | claude-fable-5, claude-opus-5/4.8/4.7/4.6, sakana/fugu-ultra |
+| Plan | Allowed | Blocked |
+|------|---------|---------|
+| `individual-go` | open source + GPT-5.6 Luna, Grok 4.5, Qwen Max & Plus | all other premium |
+| `individual-pro` | premium + opensource | anthropic:claude-fable-5, claude-opus-5/4.8/4.7/4.6, vercel-ai-gateway:sakana/fugu-ultra |
 | `individual-provider` | premium + opensource | (none) |
 | `individual-max` | premium + opensource | (none) |
 | `individual-ultra` | premium + opensource | (none) |
 | `teams-pro` | premium + opensource | (none) |
 
-Model list in TUI page 5 uses `_orderedModels` (sorted by plan). Go shows opensource first (green), premium dimmed (grey). Pro+ shows premium first.
+**Credits override:** `purchasedCredits > 0` or `freeCredits > 0` grants access
+to every model regardless of plan (matches the CLI).
+
+Model list in TUI page 5 is rendered from the **live model list** (fetched from
+`/provider/v1/models` on refresh, merged with the 52 bundled `ModelsDb` models).
+All "available on your plan" models are grouped together at the top, then
+sub-grouped by usage (Free vs Billing) and by provider. Blocked premium models
+are listed below. Go shows accessible models first (green), blocked premium
+dimmed (grey). Pro+ shows all accessible first. `PlanAccess.isAccessible`
+drives the grouping and honors the credits override. `Enter` copies the
+highlighted model (same `_orderedModels` index used for rendering).
+
+The bridge's own `/v1/models` endpoint also merges the live list so newly
+released models (e.g. `inclusionai/ling-3.0-flash-free`) are served without a
+bridge release.
 
 ## Port
 
@@ -265,7 +281,7 @@ Copy triggers:
 | `2` | Plan & Billing | `/alpha/billing/subscriptions` + `/alpha/billing/credits` | Progress bars for period elapsed, credit usage, individual credit types |
 | `3` | Usage | `/alpha/usage/summary` | Success rate bar, input/output token bars, credit breakdown bars, cost |
 | `4` | Rate Limits | `/alpha/billing/credits` (windowLimits) | 5-hour and weekly usage bars with exceed warnings, remaining, reset times |
-| `5` | Models | `_orderedModels` (sorted by plan, 44 total) | Text list with copy-on-Enter |
+| `5` | Models | Live `/provider/v1/models` merged with 52 bundled models | Text list with copy-on-Enter |
 | `6` | Proxy Config | Bridge state + endpoints | Text info |
 | `7` | Cost Sync | Local config files | Detected agents, model list with pricing, sync status |
 
@@ -292,7 +308,7 @@ Visualization uses `ProgressBar` from nocterm with color-coded fill:
 | `Ctrl+L` | Always | Toggle log sidebar |
 | `f` | Log open | Toggle log fullscreen / sidebar |
 | `Shift+C` | Log open | Clear all logs (with Y/N confirmation) |
-| `O` | Log open | Clear logs before today (with Y/N confirmation) |
+| `Shift+O` | Log open | Clear logs before today (with Y/N confirmation) |
 | `l` | Not auth'd | Open login instructions panel |
 | `Esc` | Sub-panels | Back to main |
 

CHANGELOG.md

@@ -2,6 +2,23 @@
 
 All notable changes to this project will be documented in this file.
 
+## v1.3.0 (2026-08-02)
+
+### Features
+
+- Dynamic model list: bridge now fetches the live Command Code model catalog from `/provider/v1/models` on refresh and merges it with the bundled list, so newly released models (e.g. `inclusionai/ling-3.0-flash-free`, `poolside/laguna-s-2.1-free`) appear in `/v1/models` and the TUI without a bridge release
+- Served `/v1/models` endpoint is now dynamic too: it merges the live API list with the bundled registry, so OpenAI/Anthropic clients see newly released Command Code models immediately
+- TUI page 5 (Models): renders the live model list with correct per-plan access grouping (Go/Pro/Max + credits override), showing which models are actually usable on the user's plan
+- TUI page 5 (Models): all "available on your plan" models grouped together at the top, then sub-grouped by usage (Free usage vs Billing) and by provider (DeepSeek, Moonshot, ZAI, MiniMax, Xiaomi, Qwen, StepFun, Tencent, Nvidia, Thinking Machines, OpenAI, xAI, ...); Enter-to-copy stays in sync with the highlighted row
+- TUI page 7 (Cost): pricing grouped by plan access first (Available on Go vs Requires Pro/Max), not by opensource/premium; live-known models without local pricing are shown as "no pricing data" (yellow) instead of being hidden
+- TUI page 7 (Cost): sync result now appears as a status-bar notification (green on success, red on failure) instead of being rendered at the bottom of the page where it could go unnoticed
+- Plan access model mirrors the official CLI (`evaluateModelAccess`): Go = open source + `gpt-5.6-luna`, `xai/grok-4.5`; Pro blocks Fable/Opus + Fugu Ultra; credits override grants everything
+- Expanded bundled model registry from 44 to 52 models (added Kimi-K3, Qwen 3.7 Flash, Gemini 3.6 Flash, Gemini 3.5 Flash Lite, Inkling, Inkling Small, Laguna S 2.1, Ling 3.0 Flash)
+
+### Fixes
+
+- Keymap: `o` now always copies the OpenAI endpoint URL; "Clear entries before today" moved to `Shift+O` (previously `o` in the log panel cleared old logs instead of copying the endpoint)
+
 ## v1.2.0 (2026-07-31)
 
 ### Features

README.md

@@ -14,15 +14,16 @@ APIs for use with OpenCode, Zed, Cursor, or any compatible client.
 - **OpenAI Compatible** — `/v1/chat/completions` endpoint, streaming + non-streaming
 - **Anthropic Compatible** — `/v1/messages` endpoint with full SSE streaming
 - **TUI Dashboard** — 7 info pages with progress bars and real-time log
-- **Plan-Aware** — Models sorted by plan access level
+- **Dynamic Models** — Fetches the live Command Code model catalog from `/provider/v1/models` on refresh, so newly released models appear automatically
+- **Plan-Aware** — Models grouped by what your plan can actually use (Go/Pro/Max + credits override)
 - **Cost Sync** — Sync Command Code model pricing to CLI agent configs (OpenCode, Aider, Goose)
 - **Port Config** — Change port via TUI with availability scan, persisted across restarts
 - **Cross-platform** — Linux (primary), macOS/Windows (experimental)
 
 ## Quick Install
 
 ```
-npm install -g ./commandcode-bridge-v1.2.0.tgz
+npm install -g ./commandcode-bridge-v1.3.0.tgz
 commandcode-bridge run
 ```
 

docs/API-REFERENCE.md

@@ -9,7 +9,7 @@ Base URL: `http://127.0.0.1:17077/v1`
 | Path | Method | Description |
 |------|--------|-------------|
 | `/v1/chat/completions` | POST | Chat completions (stream + non-stream) |
-| `/v1/models` | GET | List 44 available models |
+| `/v1/models` | GET | List available models (live API merged with bundled, 52+) |
 | `/v1/health` | GET | Health check |
 | `/v1/token` | GET | Get current access token |
 | `/v1/info` | GET | Bridge info and config |

docs/ARCHITECTURE.md

@@ -22,13 +22,13 @@ commandcode-bridge/
 │       ├── main.dart                  # CLI wiring (TUI / server / cost-sync modes)
 │       ├── models/
 │       │   ├── account.dart           # Account + config store (port persist)
-│       │   ├── models_db.dart         # 44 models with plan access metadata
+│       │   ├── models_db.dart         # 52 bundled models + plan access metadata
 │       │   └── version.dart           # Bridge version constant
 │       ├── services/
-│       │   ├── api_client.dart        # HTTP client for api.commandcode.ai
+│       │   ├── api_client.dart        # HTTP client for api.commandcode.ai (incl. /provider/v1/models)
 │       │   ├── cost_sync.dart         # Cost sync: detect agents, update configs
 │       │   ├── log_store.dart         # JSONL activity log (2000 entries max)
-│       │   ├── pricing_db.dart        # Hardcoded pricing table (44 models)
+│       │   ├── pricing_db.dart        # Hardcoded pricing table (52 models)
 │       │   └── updater.dart           # Self-update: API cache + download .tgz + npm install -g
 │       ├── server/
 │       │   ├── server_controller.dart  # HTTP server, routing, shared endpoints
@@ -68,7 +68,7 @@ Client -> POST /v1/messages           -> Same flow, Anthropic format
 The `cost-sync` CLI command and TUI page 7 keep CLI agent cost tracking in
 sync with Command Code pricing.
 
-- `pricing_db.dart` holds a hardcoded pricing table (44 models) matching the
+- `pricing_db.dart` holds a hardcoded pricing table (52 models) matching the
   live bridge `/v1/models` API
 - `cost_sync.dart` detects installed CLI agents (OpenCode, Aider, Goose),
   reads each agent's configured models, and writes per-model cost fields
@@ -78,6 +78,23 @@ sync with Command Code pricing.
 - Pricing is validated against the live `/v1/models` endpoint before syncing
 - OpenCode configs (JSONC) are parsed with comment and trailing-comma support
 
+## Dynamic Model List
+
+The bridge does not rely solely on its bundled registry. On every TUI refresh
+(`[r]` or auto-refresh) it calls the official Command Code endpoint
+`GET /provider/v1/models`, caches the result in memory, and merges it with the
+52 bundled `ModelsDb` models. This means newly released models (for example
+`inclusionai/ling-3.0-flash-free`, `poolside/laguna-s-2.1-free`) are served by
+`/v1/models` and shown in the TUI without needing a bridge release.
+
+- `api_client.fetchModels()` returns the live model list from the API
+- `ServerController.setLiveModelIds()` updates the in-memory cache used by `/v1/models`
+- Unknown live models are still proxyable and appear with `owned_by: command-code`
+- TUI page 5 groups models by plan access (Go/Pro/Max) using `PlanAccess`, which
+  mirrors the official CLI's `evaluateModelAccess`: Go = open source + GPT-5.6
+  Luna / Grok 4.5, Pro blocks Fable/Opus + Fugu Ultra, and a credits override
+  (purchased or free credits) grants access to everything
+
 ## Protocol Translation
 
 ### OpenAI Compatible

docs/TUI.md

@@ -8,7 +8,7 @@
 | `2` | Plan & Billing | `/alpha/billing/subscriptions` + credits | Progress bars for billing period, credit usage |
 | `3` | Usage | `/alpha/usage/summary` | Success rate bar, token bars, cost breakdown |
 | `4` | Rate Limits | `/alpha/billing/credits` | 5-hour and weekly usage bars with exceed warnings |
-| `5` | Models | 44 models sorted by plan access | Scrollable list, Enter to copy codename |
+| `5` | Models | Live API + 52 bundled models, sorted by plan access | Scrollable list, Enter to copy codename |
 | `6` | Proxy Config | Bridge state | Port, API URL, endpoints, uptime |
 | `7` | Cost Sync | Local CLI agent configs | Detected agents, provider list, model pricing, sync status |
 
@@ -31,7 +31,7 @@
 | `Ctrl+L` | Always | Toggle log sidebar |
 | `f` | Log open | Toggle log fullscreen / sidebar |
 | `Shift+C` | Log open | Clear all logs (with Y/N confirmation) |
-| `O` | Log open | Clear old entries (with Y/N confirmation) |
+| `Shift+O` | Log open | Clear old entries (with Y/N confirmation) |
 | `l` | Not authenticated | Open login instructions |
 
 ## Status Notifications
@@ -56,28 +56,52 @@ The switch is logged and the running port is shown in the header.
 
 ## Plan Access
 
-Models in page 5 are prioritized by your plan:
+Page 5 renders the **live model list** (fetched from `/provider/v1/models` on
+refresh, merged with 52 bundled models) and groups models by what your plan can
+actually use. Rules mirror the official Command Code CLI:
 
-| Plan | Sort Order | Accessible |
-|------|-----------|------------|
-| individual-go | Opensource first (green), Premium dimmed | Opensource only |
-| individual-pro | Premium first | Premium + opensource (blocked: top Claude + Fugu) |
-| individual-max | Premium first | All models |
-| individual-ultra | Premium first | All models |
-| teams-pro | Premium first | All models |
+| Plan | Accessible |
+|------|-----------|
+| individual-go | Open source models + GPT-5.6 Luna, Grok 4.5, Qwen Max & Plus |
+| individual-pro | Premium + opensource, blocked: Claude Fable 5, Opus 5/4.8/4.7/4.6, Fugu Ultra |
+| individual-provider / max / ultra / teams-pro | All models |
+
+**Credits override:** if you have purchased or free credits
+(`purchasedCredits > 0` or `freeCredits > 0`), every model becomes accessible
+regardless of plan.
+
+### Page 5 Layout
+
+Models are grouped top-level by plan access first (all "available on your
+plan" models together at the top), then:
+
+1. **Usage:** free models (`Free usage`) before billed models (`Billing`)
+2. **Provider:** within Billing, sub-grouped by provider (DeepSeek, Moonshot,
+   ZAI, MiniMax, Xiaomi, Qwen, StepFun, Tencent, Nvidia, Thinking Machines,
+   OpenAI, xAI, ...)
+
+Blocked premium models are listed below under their own header. Use `up/down`
+to select a model and `Enter` to copy its ID; the copy always matches the
+highlighted row.
 
 ### Go Plan Accessible Models
 
+Open source models plus the premium exceptions included on Go:
+GPT-5.6 Luna, Grok 4.5, and Qwen Max & Plus. The rest of the premium lineup
+(Claude, other GPT, Gemini, Muse Spark) is blocked on Go.
+
 - DeepSeek V4 Pro, DeepSeek V4 Flash
+- Kimi K3, K2.7 Code, K2.7 Code HighSpeed, K2.6, K2.5
 - MiniMax M3 Free, M3, M2.7, M2.5
-- Kimi K2.7 Code, K2.6, K2.5
-- Qwen 3.7 Max/Plus, 3.6 Max Preview/Plus
+- Qwen 3.7 Max/Plus/Flash, 3.6 Max Preview/Plus
 - GLM 5.2, 5.2 Fast, 5.1, 5
 - MiMo V2.5 Pro, V2.5
 - Step 3.7 Flash, 3.5 Flash
 - Tencent HY3 Paid, HY3
-- Meta Muse Spark 1.1
+- Thinking Machines Inkling, Inkling Small
+- Poolside Laguna S 2.1 Free (Free usage), InclusionAI Ling 3.0 Flash Free (Free usage)
 - Nvidia Nemotron 3 Ultra
+- GPT-5.6 Luna, Grok 4.5
 
 ## Cost Sync (Page 7)
 
@@ -86,11 +110,17 @@ agent-side cost tracking shows real per-token rates.
 
 - `[7]` opens the Cost Sync page
 - `[c]` syncs pricing for the selected agent (hint shown on the page)
+- Sync progress and result appear as a status-bar notification: green on success
+  (e.g. "Cost sync OpenCode: 5 updated, 2 already configured"), red if any
+  models failed, so the result is always visible without scrolling the page
 - `[up/down]` selects the agent (OpenCode, Aider, Goose)
 - Detected agents are listed with their config paths
 - Bridge providers are listed by name and endpoint, for example
   `Khip01 - Command Code (Go Plan) [127.0.0.1:17077]`
-- All Command Code models with pricing are shown, grouped by Open Source and Premium
+- All Command Code models with pricing are shown, grouped by plan access first
+  ("Available on Go" / "Requires Pro / Max", or "Available on your plan" /
+  "Blocked on your plan" for other plans)
+- Live-known models without local pricing data appear as "no pricing data" (yellow)
 - Models in your config appear bright white with "(will be implemented)"
 - Models not in your config are greyed out
 - Pricing is validated against the live bridge `/v1/models` before syncing