Add Ollama and llama.cpp local LLM providers to SDK

[brainless]

Summary

This PR adds support for local LLM providers to the nocodo-llm-sdk, enabling developers to use locally-hosted models through Ollama and llama.cpp. This expands the SDK's capabilities beyond cloud-based providers, offering privacy, cost savings, and offline functionality.

Key Additions

• Ollama Provider: Full integration with Ollama's /api/chat endpoint

  • Support for local models (Llama 3.1, Ministral, etc.)
  • Tool/function calling support
  • System prompt handling
  • Configurable base URL (defaults to http://localhost:11434)
  • Debug logging for requests/responses via NOCODO_LLM_LOG_PAYLOADS

• llama.cpp Provider: OpenAI-compatible API integration

  • Supports llama.cpp server mode
  • Function calling capabilities
  • Configurable base URL (defaults to http://localhost:8080)

• Documentation: Added comprehensive README examples and external API documentation

Bug Fixes (Ollama)

• Fixed system prompt being ignored in requests
• Fixed streaming issues by explicitly setting stream=false
• Removed $schema from tool parameters (causes Ollama validation errors)
• Added debug logging for troubleshooting

Files Changed

• 16 files changed, 2,062 insertions
• New modules: ollama/ and llama_cpp/ with builder, client, types, and tools
• Updated models and providers list
• Added external API documentation for reference

Test Plan

• Test Ollama integration with a local Llama model
• Test llama.cpp integration with llama-server
• Verify function calling works with both providers
• Test system prompts are correctly handled
• Verify debug logging with NOCODO_LLM_LOG_PAYLOADS=1
• Confirm default URLs work for both providers
• Test custom base URLs for both providers

🤖 Generated with Claude Code

[github-actions]

📊 Code Complexity Analysis

• Lines added: 2062
• Lines removed: 1
• Net change: 2061

💡 Suggestion: This is a large PR with 2062 added lines. Consider:

• Breaking it into smaller, focused PRs
• Adding comprehensive tests for the new functionality
• Updating documentation as needed

---

Automated analysis by GitHub Actions

[github-actions]

🤖 Automated Code Review Summary

This automated review was generated to help ensure code quality and security standards.

Rust Code Analysis

• ⚠️ Code formatting: Some Rust files are not formatted according to rustfmt standards.

  • Run cargo fmt to fix formatting issues.

• ⚠️ Linting: Clippy found potential issues in Rust code.

  • Run cargo clippy --workspace --all-targets -- --deny warnings to see detailed warnings.

Security Analysis

• ⚠️ Potential secrets: Found references to passwords, secrets, or tokens.

  • Please verify no hardcoded credentials are present.

• ℹ️ Debug output: Found debug print statements.

  • Consider removing or replacing with proper logging.

Recommendations

• Run the full CI pipeline to ensure all tests pass
• Consider adding tests for any new functionality
• Update documentation if API changes are involved
• Follow the development workflow described in CLAUDE.md

---

This review was automatically generated. Please address any issues before merging.