[codex] Add OpenAI-compatible LLM providers

[jsdavid278-cyber]

Summary

• Add a shared OpenAI-compatible agent provider factory for chat completions and model listing.
• Add Groq, DeepSeek, Mistral, xAI, and Cerebras provider packages using that shared factory.
• Add focused provider metadata/env tests and pnpm workspace lockfile entries.

Why

This fills the requested OpenAI-compatible LLM adapter set for Groq, DeepSeek, Mistral, xAI, and Cerebras while keeping provider-specific packages small and consistent with existing agent-provider package structure.

Validation

• Passed: tsc -p packages/agent-providers/shared/tsconfig.json --noEmit
• Passed: tsc -p packages/agent-providers/groq/tsconfig.json --noEmit
• Passed: tsc -p packages/agent-providers/deepseek/tsconfig.json --noEmit
• Passed: tsc -p packages/agent-providers/mistral/tsconfig.json --noEmit
• Passed: tsc -p packages/agent-providers/xai/tsconfig.json --noEmit
• Passed: tsc -p packages/agent-providers/cerebras/tsconfig.json --noEmit
• Note: focused Vitest run could not start locally because Vite/esbuild hit Windows spawn EPERM; TypeScript package checks above passed.

[greptile-apps]

Greptile Summary

This PR introduces a shared createOpenAICompatibleProvider factory and five new provider packages (Groq, DeepSeek, Mistral, xAI, Cerebras) that use it. All default models were verified against official API documentation and are current and valid.

• Shared factory (openai-compatible.ts): handles chat completions, model listing, env validation, URL normalization, and structured error-body extraction — well-tested with a dedicated test suite.
• Five provider packages: each is a thin config wrapper around the shared factory, consistent with the existing openrouter package structure; base URLs and default models confirmed correct.
• pnpm lockfile: updated with workspace entries for all five new packages.

Confidence Score: 5/5

All five providers are thin wrappers around a well-tested shared factory; no logic errors or API contract violations detected.

The shared factory correctly handles URL normalization, auth headers, error-body extraction, and env validation. All default model IDs were cross-checked against official provider documentation and are confirmed valid. The package structure mirrors the existing openrouter provider. The change is purely additive with no modifications to existing code paths.

No files require special attention.

Important Files Changed

Filename	Overview
packages/agent-providers/shared/src/openai-compatible.ts	New shared factory for OpenAI-compatible providers; implements chat, listModels, validateEnv, and error detail extraction correctly
packages/agent-providers/shared/src/tests/openai-compatible.test.ts	Good coverage of metadata, env validation, chat, listModels, and error body propagation through the factory
packages/agent-providers/groq/src/provider.ts	Thin wrapper around shared factory; correct base URL and valid default model (llama-3.3-70b-versatile)
packages/agent-providers/deepseek/src/provider.ts	Valid base URL and confirmed current model (deepseek-v4-flash per official DeepSeek API docs as of 2026)
packages/agent-providers/mistral/src/provider.ts	Correct base URL and valid alias model (mistral-large-latest)
packages/agent-providers/xai/src/provider.ts	Valid base URL and confirmed current model (grok-4.3 released April 2026)
packages/agent-providers/cerebras/src/provider.ts	Valid base URL and confirmed model (gpt-oss-120b is a production Cerebras model per inference-docs.cerebras.ai)
packages/agent-providers/cerebras/package.json	Consistent with existing openrouter package structure; main points to TS source for workspace resolution, dist for publish

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[Caller] --> B[Provider Package\ne.g. groqProvider]
    B --> C[createOpenAICompatibleProvider\nshared factory]
    C --> D{Operation}
    D --> E[chat\nPOST /chat/completions]
    D --> F[listModels\nGET /models]
    D --> G[validateEnv\ncheck API key + URL]
    D --> H[healthcheck\nvalidateEnv only]
    E --> I{res.ok?}
    F --> I
    I -->|Yes| J[parseContent / sort IDs]
    I -->|No| K[readErrorDetail\nres.text → JSON.parse → message]
    K --> L[AgentProviderRequestError\ndisplayName + op + status + message]
    J --> M[Return result]

    subgraph Providers
        P1[groq]
        P2[deepseek]
        P3[mistral]
        P4[xai]
        P5[cerebras]
    end
    Providers --> C

Loading

_{Reviews (2): Last reviewed commit: "Add OpenAI-compatible LLM providers" | Re-trigger Greptile}

[github-actions]

🤖 Auto-rebase: The branch was rebased successfully locally but could not be pushed to the fork. Please enable 'Allow edits from maintainers' in the PR settings, or rebase manually: git fetch upstream master && git rebase upstream/master.

[github-actions]

🤖 Auto-rebase: The branch was rebased successfully locally but could not be pushed to the fork. Please enable 'Allow edits from maintainers' in the PR settings, or rebase manually: git fetch upstream master && git rebase upstream/master.

[github-actions]

🤖 Auto-rebase: The branch was rebased successfully locally but could not be pushed to the fork. Please enable 'Allow edits from maintainers' in the PR settings, or rebase manually: git fetch upstream master && git rebase upstream/master.