Skip to content

github-models.md

Latest commit

 

History

History
286 lines (194 loc) · 9.06 KB

github-models.md

File metadata and controls

286 lines (194 loc) · 9.06 KB

GitHub Models

📌 Important Note

Before configuring, please review:

GitHub Models is separate from GitHub Copilot UI features.

  • GitHub Copilot: IDE/chat coding assistant experience
  • GitHub Models: REST API access to hosted models via https://models.github.ai

aicommit2 integrates with the official GitHub Models API, not private Copilot endpoints.

🚀 Quick Setup

Option 1: Automatic login (recommended)

aicommit2 github-login

This command:

  1. Authenticates with GitHub CLI
  2. Stores token in GITHUB_MODELS.key
  3. Verifies GitHub Models access

Option 2: Manual token setup

  1. Create a GitHub Personal Access Token (PAT)
  2. Grant permission: models: read
  3. Configure token:
aicommit2 github-login --token github_pat_xxxxxxxxxxxxxxxxxxxx
# or
aicommit2 config set GITHUB_MODELS.key="github_pat_xxxxxxxxxxxxxxxxxxxx"

PAT Generation (Important)

Create a token from GitHub settings:

You can find additional information at the link: GitHub personal access tokens documentation.

Token must have models: read access, otherwise requests typically fail with 403.

models: read is for GitHub Models API usage. It is different from Copilot CLI/SDK token requirements (Copilot Requests).

Supported token prefixes in aicommit2 include:

  • github_pat_...
  • ghp_...
  • gho_...
  • ghu_...
  • ghs_...
  • ghr_...

Prerequisites (for github-login)

  • GitHub CLI installed
  • Browser access for web login

Install GitHub CLI:

# macOS
brew install gh

# Windows (winget)
winget install GitHub.cli

# Windows (choco)
choco install gh

# Linux (Debian/Ubuntu)
curl -fsSL https://cli.github.com/packages/githubcli-archive-keyring.gpg | sudo dd of=/usr/share/keyrings/githubcli-archive-keyring.gpg
echo "deb [arch=$(dpkg --print-architecture) signed-by=/usr/share/keyrings/githubcli-archive-keyring.gpg] https://cli.github.com/packages stable main" | sudo tee /etc/apt/sources.list.d/github-cli.list > /dev/null
sudo apt update
sudo apt install gh

Model ID Format (Required)

GITHUB_MODELS.model must use this format:

publisher/model

Examples:

  • openai/gpt-4o-mini
  • openai/gpt-5
  • openai/gpt-5-chat
  • meta/llama-3.3-70b-instruct

Legacy short IDs like gpt-4o-mini are invalid for this provider flow in aicommit2.

Copilot Plan Limits (Important)

Model access depends on your GitHub Copilot plan and request allowance.

As of April 2, 2026 (based on current GitHub docs):

  • Copilot Free: up to 2,000 inline suggestion requests and up to 50 premium requests/month. Model choice is limited.
  • Paid plans / Copilot Student: unlimited inline suggestions and unlimited chat interactions for included models (GPT-5 mini, GPT-4.1, GPT-4o), with additional premium request allowances for premium models.
  • GitHub explicitly notes model availability can vary by plan and can change over time.
  • Additional premium request purchases are not available on Copilot Free.

You can find additional information at the link: Plans for GitHub Copilot and Requests in GitHub Copilot.

Important for aicommit2 users

Even if a model appears available in Copilot UI, it may still be unavailable for your token on models.github.ai.

For API usage with aicommit2, you may see:

  • 400 unavailable_model (model not currently usable for your token in API context)
  • 403 (permission/entitlement issue)
  • 429 (rate limiting)

When in doubt, run a live check with your own token before setting a default model.

Validated Snapshot (April 2, 2026)

The following results were validated with a real GITHUB_TOKEN on 2026-04-02 using:

  • X-GitHub-Api-Version: 2026-03-10
  • catalog probe + throttled dry-run commit generation

Probe result (catalog text models)

  • Total probed models: 41
  • 200 OK: 29
  • 400: 7 (unavailable_model / unknown_model)
  • 403: 5

Dry-run commit generation result (for 200 OK models)

  • Total dry-run models: 29
  • OK (generated): 23
  • FAIL (no_valid_commit_message): 6

Models that generated valid commit messages (OK)

  • OpenAI: openai/gpt-4.1, openai/gpt-4.1-mini, openai/gpt-4.1-nano, openai/gpt-4o, openai/gpt-4o-mini
  • Cohere: cohere/cohere-command-a, cohere/cohere-command-r-08-2024, cohere/cohere-command-r-plus-08-2024
  • DeepSeek: deepseek/deepseek-v3-0324
  • Meta: meta/llama-3.2-11b-vision-instruct, meta/llama-3.2-90b-vision-instruct, meta/llama-3.3-70b-instruct, meta/llama-4-maverick-17b-128e-instruct-fp8, meta/llama-4-scout-17b-16e-instruct, meta/meta-llama-3.1-405b-instruct, meta/meta-llama-3.1-8b-instruct
  • Mistral AI: mistral-ai/codestral-2501, mistral-ai/ministral-3b, mistral-ai/mistral-medium-2505, mistral-ai/mistral-small-2503
  • xAI: xai/grok-3
  • Microsoft: microsoft/phi-4, microsoft/phi-4-multimodal-instruct

Models that responded but failed commit JSON validation (FAIL)

  • deepseek/deepseek-r1
  • deepseek/deepseek-r1-0528
  • xai/grok-3-mini
  • microsoft/phi-4-mini-instruct
  • microsoft/phi-4-mini-reasoning
  • microsoft/phi-4-reasoning

Important interpretation

  • This snapshot is token-specific and time-specific.
  • A model can appear in catalog/models and still fail at inference time for your token (unavailable_model, 403, etc.).
  • A model can return 200 but still fail aicommit2 commit parsing if it does not return valid commit JSON.
  • This is currently the most reliable way to find working models for your account: probe + throttled dry-run.

Reproduce safely (low rate-limit risk)

  1. Fetch catalog with X-GitHub-Api-Version: 2026-03-10.
  2. Probe only text-generation models with a delay between requests (for example, sleep 2).
  3. Keep only models with HTTP 200.
  4. Run commit --dry-run checks sequentially with throttling (for example, sleep 12) and retry on 429.

Example catalog command:

curl -sSL \
  -H "Accept: application/vnd.github+json" \
  -H "Authorization: Bearer $GITHUB_TOKEN" \
  -H "X-GitHub-Api-Version: 2026-03-10" \
  https://models.github.ai/catalog/models > catalog.json

Usage Examples

Basic usage

aicommit2 github-login
aicommit2 config set GITHUB_MODELS.model="openai/gpt-5"

git add .
aicommit2

Advanced configuration

aicommit2 config set \
  GITHUB_MODELS.key="github_pat_xxxxxxxxxxxxxxxxxxxx" \
  GITHUB_MODELS.model="openai/gpt-5" \
  GITHUB_MODELS.temperature=0.7 \
  GITHUB_MODELS.maxTokens=1024 \
  GITHUB_MODELS.locale="en" \
  GITHUB_MODELS.generate=3 \
  GITHUB_MODELS.topP=0.95

Settings

Setting Description Default
key GitHub token -
model GitHub Models model ID (publisher/model) openai/gpt-4o-mini

Discover Available Models

GitHub catalog changes over time. Use live discovery instead of static hardcoded lists.

# Install extension once
gh extension install https://github.com/github/gh-models

# Show available models
gh models list

# Optional: raw catalog endpoint
curl -L \
  -H "Accept: application/vnd.github+json" \
  -H "Authorization: Bearer $GITHUB_TOKEN" \
  -H "X-GitHub-Api-Version: 2026-03-10" \
  https://models.github.ai/catalog/models

You can find additional information at the link: GitHub Models inference REST API.

API Version

aicommit2 uses GitHub API version header:

X-GitHub-Api-Version: 2026-03-10

Troubleshooting

401 Unauthorized

Token is invalid/expired:

aicommit2 github-login

403 Forbidden

Token lacks models: read permission.

404 Model not found

Model unavailable for your account or incorrect model ID. Check gh models list.

422 Validation error

Model ID format is invalid. Use publisher/model (example: openai/gpt-4o-mini).

References

You can find additional information at the link: