release: v0.4.0

[code-crusher]

v0.4.0

Added

• /task slash command — reference a previous session as context for the current task
• axon-eido-3-flash — new free model replacing axon-code-2-5-mini (200K context, /bin/zsh)

Changed

• System prompt rewritten for speed and editing discipline — gather enough context, then act principle, editing discipline rules, reading/search hygiene, and a plan-before-editing mandate

Fixed

• Transient stream failures auto-retried — up to 3 retries with exponential backoff for connection drops and 5xx/408/429
• Reasoning timer now shows only thinking time — open/close segments instead of a single flag
• Mid-stream retry allowed when partial output can be rolled back

---

Per the release process in RELEASE.md: after this PR is merged into main, tag and push v0.4.0 to trigger the npm publish.

[matterai-app]

Summary By MatterAI

🔄 What Changed

This PR marks the release of version v0.4.0. It includes a version bump in package.json and the addition of the axon-eido-3-flash model to the built-in model registry. This new model is configured as a free, high-performance option with a 200,000 token context window.

🔍 Impact of the Change

Users now have access to a zero-cost, fast, general-purpose model (axon-eido-3-flash) for low-effort tasks. The update also formalizes the project versioning to 0.4.0, ensuring consistency across the CLI and API interactions.

📁 Total Files Changed

Click to Expand

File	ChangeLog
Version Bump package.json	Updated project version from previous to 0.4.0.
Model Registry src/api/models.ts	Added axon-eido-3-flash configuration with pricing and capability metadata.

🧪 Test Added/Recommended

Recommended

• Integration Test: Verify that the CLI correctly identifies and routes requests to the new axon-eido-3-flash model.
• Validation Test: Ensure that the inputPrice and outputPrice of 0.0 do not cause division-by-zero errors in any cost-calculation logic.

🔒 Security Vulnerabilities

No direct security vulnerabilities detected. The changes are limited to configuration and versioning metadata. 🚀

⏳ Estimated code review effort

LOW (~5 minutes)

Tip

Quality Recommendations

1. Ensure that the backend gateway is synchronized to support the new 'axon-eido-3-flash' model ID before deployment.

2. Consider adding a unit test to validate that all models in BUILTIN_AXON_MODELS adhere to the AxonModel interface strictly.

♫ Tanka Poem

A flash of new light ⚡
Zero cost for every soul 🧪
Context wide and deep 🌊
Version four is now alive
Code flows like a mountain stream 🏔️

Sequence Diagram

sequenceDiagram
    participant CLI as OrbCode CLI
    participant Registry as Model Registry (models.ts)
    participant Gateway as MatterAI Gateway

    Note over CLI, Registry: User selects model
    CLI->>Registry: getModel("axon-eido-3-flash")
    Registry-->>CLI: return AxonModel { free: true, context: 200k }

    Note over CLI, Gateway: Execution Flow
    CLI->>Gateway: POST /chat/completions { model: "axon-eido-3-flash" }
    Gateway-->>CLI: AI Response

Loading

[matterai-app]

✅ Reviewed the changes: Release v0.4.0: version bump and model registry update replacing axon-code-2-5-mini with the new axon-eido-3-flash free model. No issues found in the changed lines.