release: v6.8.0

CHANGELOG.md

@@ -9,6 +9,42 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Added
 
+- **Dynamic model catalog synchronization.** OrbCode now fetches the active model catalog dynamically from the backend (`/v1/models`) on startup and when refreshing usage (`fetchDynamicModels`), registering returned OSS models into `BUILTIN_AXON_MODELS` and `AXON_MODELS` so newly added models appear in the picker without requiring hardcoded updates. Models the backend retires are pruned after a successful fetch (empty or failed responses never wipe the offline fallback), and the catalog's `iconUrl` / `costMultiplier` fields are captured on each model.
+- **Provider badges in the model picker.** Terminals can't render the catalog's SVG provider icons, so picker rows show a text badge (`[Z.ai]`, `[Meta]`, `[DeepSeek]`, `[OpenAI]`, `[Google]`) — the TUI equivalent of the webapp's provider logos.
+- **`orbcode usage` command.** Prints the weekly/monthly plan usage windows
+  (percentage bars with reset times) and each tracked OSS model's share of
+  the shared plan pool as weekly/monthly percentages, alongside the model's
+  plan-cost multiplier (e.g. `5x cost`). The TUI's `/usage` and `/status`
+  commands show the same per-model block. Percentages only — no credit
+  amounts are exposed. Requires a logged-in token (`orbcode login`).
+
+### Changed
+
+- **Built-in model catalog is now OSS-first.** The built-in registry replaces
+  the Axon models with seven OSS models served through the MatterAI gateway:
+  `zai/glm-5.3-flash` (the new default), `zai/glm-5.3`,
+  `deepseek/deepseek-v4-flash-0731`, `meta/muse-spark-1.3-contributor`,
+  `gpt-5.6-luna`, `gpt-5.6-sol`, and `gemini-3.8-flash`. All seven expose a
+  232K context window with 64K max output, are available on every plan, and
+  carry their published per-token pricing. The 400K
+  context variants and `axon-auto` are gone from the picker; a stored Axon
+  model selection auto-resets to the new default on next launch, and a
+  requested Axon id (`--model` / `MATTERAI_MODEL`) now warns and falls back
+  to the default.
+
+## [6.8.0] - 2026-08-28
+
+### Changed
+
+- **Ported the 6.8.2 coding-harness update from the Orbital extension.**
+  - `search_files` is now one-shot: ripgrep-first with FFF fallback, results bounded to the first 100 matches (default `max_results` 100), and cursor pagination removed from the model-facing schema and output. Capped results tell the model to refine the query instead of paginating.
+  - Independent read-only tool calls (`read_file`, `search_files`, `list_files`, `list_code_definition_names`, `codebase_search`, `lsp`) at the start of an assistant response now execute concurrently (max 4) with results committed in model order; mutating and interactive tools stay serialized.
+  - Malformed tool-call JSON now returns a corrective tool result that includes the raw arguments, so the model can re-issue the call with valid JSON instead of dead-ending.
+  - Native tool schemas tightened for strict mode: optional parameters are now required with nullable types (`replace_all`, `recursive`, `follow_up`, `offset`/`limit`, `cwd`/`message`/`isDangerous`, and the inactive-in-CLI tool schemas), and `execute_command` guidance asks for an explicit safety classification.
+  - System-prompt `search_files` guidance updated to the bounded one-shot behavior.
+
+### Added
+
 - **Investigation efficiency guidance in system prompt.** Added an "Investigation efficiency" section to the tool guide (`src/prompts/system.ts`) that directs the agent to classify comprehension questions separately from implementation tasks, form a one-line hypothesis before searching, read call sites rather than implementation internals, avoid reading prose/content when the question is about control flow, and stop exploring as soon as it can answer.
 - **Zero-result guidance in `search_files`.** `searchFiles.ts` executor now appends actionable guidance when a search returns 0 matches, directing the model to tighten or simplify the regex, widen the path scope, try a different glob, or stop searching after 2+ failed attempts.
 - **Native tool description improvements.** The `read_file` schema description now tells the model not to read file contents (prompt text, config values, prose) when investigating control flow, and not to re-read regions already read earlier. The `search_files` schema description now tells the model to scope the path to the narrowest plausible directory and to stop after 2+ zero-result searches.

README.md

@@ -122,6 +122,7 @@ runtime — bumping the version there is all that's needed.
 orbcode                 start an interactive session in the current directory
 orbcode "<prompt>"      start an interactive session with an initial prompt
 orbcode login           sign in to MatterAI (browser device flow)
+orbcode usage           show plan usage windows and per-model usage
 orbcode -p "<prompt>"   run a single prompt non-interactively, print only the final response
 orbcode -p "…" --yolo   non-interactive with edits/commands auto-approved
 orbcode --model <id>    use a specific model for this run (also -m)
@@ -177,40 +178,30 @@ Sign out with `/logout` (removes the saved token).
 
 ## Models
 
-The built-in Axon models are listed below; `/model` opens a scroll-and-select
-picker (`/model <id>` still selects directly). Additional models can be
-declared via `customModels` in settings.json. The choice persists across
-sessions.
-
-| id                             | context | max output | pricing                |
-| ------------------------------ | ------- | ---------- | ---------------------- |
-| `axon-auto-232k`               | 232k    | 64k        | dynamic pricing        |
-| `axon-auto-400k`               | 400k    | 64k        | dynamic pricing        |
-| `axon-eido-3.2-flash`          | 232k    | 64k        | $0.6/M in · $1.8/M out |
-| `axon-eido-3.2-flash-400k`     | 400k    | 64k        | $0.6/M in · $1.8/M out |
-| `axon-eido-3.2-code-232k`      | 232k    | 64k        | $2/M in · $6/M out     |
-| `axon-eido-3.2-code-400k`      | 400k    | 64k        | $2/M in · $6/M out     |
-| `axon-eido-3.2-code-pro-232k`  | 232k    | 64k        | $3/M in · $9/M out     |
-| `axon-eido-3.2-code-pro-400k`  | 400k    | 64k        | $3/M in · $9/M out     |
-| `axon-lumen-4-code-232k`       | 232k    | 128k       | $5/M in · $25/M out    |
-| `axon-lumen-4-code-400k`       | 400k    | 128k       | $5/M in · $25/M out    |
-
-`axon-auto-232k` is the default. The context suffix controls OrbCode's local
-context window; requests still send the underlying base model ID to the
-MatterAI gateway. Plan gating:
-
-- **Free**: `axon-eido-3.2-flash` (232K) only.
-- **Pro**: adds `axon-eido-3.2-code-{232k,400k}` and `axon-eido-3.2-code-pro-{232k,400k}`.
-- **Pro Plus / Ultra**: adds `axon-lumen-4-code-{232k,400k}` and unlocks every 400K variant (including `axon-eido-3.2-flash-400k`).
-
-Every 400K option — `axon-eido-3.2-flash-400k` included — is gated to Pro Plus
-and Ultra. All five options support native JSON tool calls and image input.
-Cost comes from the API's usage chunks (`is_byok`-aware) and is shown in the
-status bar.
+The built-in models are listed below; `/model` opens a scroll-and-select
+picker (`/model <id>` still selects directly). The live catalog is fetched
+from the backend at startup and kept in sync while the session runs — the
+table below is the offline fallback. Additional models can be declared via
+`customModels` in settings.json. The choice persists across sessions.
+
+| id                                | context | max output | pricing                  |
+| --------------------------------- | ------- | ---------- | ------------------------ |
+| `zai/glm-5.3-flash`               | 232k    | 64k        | $0.15/M in · $0.5/M out  |
+| `zai/glm-5.3`                     | 232k    | 64k        | $1.4/M in · $4.4/M out   |
+| `deepseek/deepseek-v4-flash-0731` | 232k    | 64k        | $0.14/M in · $0.28/M out |
+| `meta/muse-spark-1.3-contributor` | 232k    | 64k        | $0.1/M in · $0.2/M out   |
+| `gpt-5.6-luna`                    | 232k    | 64k        | $0.2/M in · $1.2/M out   |
+| `gpt-5.6-sol`                     | 232k    | 64k        | $5/M in · $30/M out      |
+| `gemini-3.8-flash`                | 232k    | 64k        | $0.75/M in · $3.75/M out |
+
+`zai/glm-5.3-flash` is the default. Every model is available on every plan,
+supports native JSON tool calls and image input, and is served through the
+MatterAI gateway. Cost comes from the API's usage chunks (`is_byok`-aware)
+and is shown in the status bar.
 
 ### Other providers (Anthropic, OpenAI-compatible)
 
-The Axon models go through the MatterAI gateway as before. A `customModels`
+The built-in models go through the MatterAI gateway as before. A `customModels`
 entry that sets a `provider` is instead served through the
 [Vercel AI SDK](https://sdk.vercel.ai), reusing the same agent loop, tools, and
 approvals — auth is the provider's own key (env var or `apiKey`), not the
@@ -412,7 +403,7 @@ Two kinds of files under `~/.orbcode/`:
 ```
 
 All keys are optional. `customModels` entries appear in the `/model` picker
-alongside the built-in Axon models; `baseUrl` points the chat client at any
+alongside the built-in models; `baseUrl` points the chat client at any
 OpenAI-compatible gateway; `env` is applied to the process at startup; `hooks`
 configures lifecycle hooks (see [Hooks](#hooks)). Precedence: env vars > project
 settings.json > user settings.json > config.json.

package.json

@@ -1,6 +1,6 @@
 {
   "name": "@matterailab/orbcode",
-  "version": "6.7.9",
+  "version": "6.8.0",
   "description": "OrbCode CLI — agentic coding in your terminal, powered by Axon models by MatterAI",
   "type": "module",
   "bin": {