Repository navigation
Releases
Releases · jhfarhanyt20-coder/VibeCoder-AI
0.3.22
release: v0.3.22 (0.3.21 features; tests read sources line-ending safe) v0.3.21's release run failed before publishing: two new tests matched "\n}\n" in source read from a CRLF checkout on the Windows runner. They now normalise line endings. Same features as the 0.3.21 commit; no release was published for v0.3.21 (like v0.3.15), so users go 0.3.20 -> 0.3.22. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
0.3.20
v0.3.20 release: read while the agent writes, no wobble, full Facebook Page c…
0.3.19
v0.3.19 release: smooth chat, code editor panel, side panel, video tools and …
0.3.18
v0.3.18 release: free-model daily limit says when it comes back, in your own …
0.3.17
v0.3.17 release: a professional CLI, a steady chat and smarter Computer Use (…
0.3.16
release: Crew workspaces work under Windows short folder names (v0.3.16) v0.3.15's release build failed: when the temp or project folder is reached through an 8.3 short name (C:\Users\RUNNER~1), git's long top-level path and the short project path gave a wrong relative path. Both are now resolved to one real spelling. Includes everything from v0.3.15 (Crew own workspaces, Gemini 3 tool calls, model pick and provider panel fixes, chat drag-select). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
0.3.14
v0.3.14 release: Computer Use, Connectors, smarter tools and a redesigned wor…
0.3.13
v0.3.13 v0.3.13: Sign in with VibeCoder and free VibeCoder models
0.3.12 - Fallback model and provider upgrades
VibeCoder AI 0.3.12
Fallback model, provider upgrades, and AgentRouter compatibility.
Fallback model
- Build one ordered chain from models across different providers.
- On a rate limit, server error, timeout, unreachable provider, or a provider content filter, the next model answers automatically.
- The cursor is sticky: it moves down the chain and wraps back to #1 at the end. It is remembered across app restarts.
- Reset to #1 is available at any time.
- Each model uses its own provider key, context window, output budget, and pricing.
- The context ring and automatic compaction follow the model that is actually answering.
Provider upgrades
- Popular providers now stay in Popular providers and turn green when configured; only custom providers appear under Custom providers.
- Ten ready-to-fill presets: OpenAI, Anthropic, DeepSeek, OpenRouter, Vercel AI Gateway, Kilo Gateway, AgentRouter, Google Gemini, Groq, and Agnes AI.
- Switch providers faster: no full catalog reload on every draft/UI write, no duplicate draft saves, and the detail panel opens at the top.
- Request headers and extra body fields are still available under Advanced (optional).
Gateway compatibility
- Every request identifies the app first (
VibeCoderAI/<version>). - If a gateway rejects the client name with
401 unauthorized client, the app retries once with a compatible client string while keeping the product header. - If a model only exists on the OpenAI Responses endpoint, the app discovers that and retries automatically.
- Clear message when a provider's content filter blocks a message.
Update
Existing installations: Settings → App Updates → Check for Updates → Update Now → Install & Restart.
New install: https://vibecoderai.onrender.com/download-page
Assets
VibeCoderAI-v2-0.3.12.exe— Windows 10/11 x64 installerVibeCoderAI-v2-0.3.12.exe.blockmaplatest.ymlVibeCoder-CLI-0.3.12.tgz— standalone CLI
Full test suite green: 125 core tests plus agents, skills, harness, control, CLI, IPC, UI regressions, and updater tests.
0.3.11
v0.3.11 release: v0.3.11 full-screen CLI, provider manager and website CLI page