This fork adds extended provider support to claude-mem.

Allows using Gemini API proxies or alternative endpoints.

{
  "CLAUDE_MEM_PROVIDER": "gemini",
  "CLAUDE_MEM_GEMINI_API_KEY": "your-api-key",
  "CLAUDE_MEM_GEMINI_BASE_URL": "https://your-proxy.com/v1beta/models",
  "CLAUDE_MEM_GEMINI_MODEL": "gemini-2.5-flash"
}

Note: Model validation is removed - any model name is accepted for proxy compatibility.

Supports any OpenAI-compatible API endpoint:

• OpenAI Official (GPT-4o, GPT-4o-mini)
• Azure OpenAI
• Local Models (Ollama, LM Studio, vLLM)
• Other Providers (Groq, Together AI, Anyscale)

Settings in ~/.claude-mem/settings.json:

{
  "CLAUDE_MEM_PROVIDER": "openai-compatible",
  "CLAUDE_MEM_OPENAI_COMPATIBLE_API_KEY": "your-api-key",
  "CLAUDE_MEM_OPENAI_COMPATIBLE_BASE_URL": "https://api.openai.com/v1/chat/completions",
  "CLAUDE_MEM_OPENAI_COMPATIBLE_MODEL": "gpt-4o-mini",
  "CLAUDE_MEM_OPENAI_COMPATIBLE_MAX_CONTEXT_MESSAGES": "20",
  "CLAUDE_MEM_OPENAI_COMPATIBLE_MAX_TOKENS": "100000"
}

Ollama (Local):

{
  "CLAUDE_MEM_PROVIDER": "openai-compatible",
  "CLAUDE_MEM_OPENAI_COMPATIBLE_API_KEY": "ollama",
  "CLAUDE_MEM_OPENAI_COMPATIBLE_BASE_URL": "http://localhost:11434/v1/chat/completions",
  "CLAUDE_MEM_OPENAI_COMPATIBLE_MODEL": "llama3.2"
}

Azure OpenAI:

{
  "CLAUDE_MEM_PROVIDER": "openai-compatible",
  "CLAUDE_MEM_OPENAI_COMPATIBLE_API_KEY": "your-azure-key",
  "CLAUDE_MEM_OPENAI_COMPATIBLE_BASE_URL": "https://your-resource.openai.azure.com/openai/deployments/your-deployment/chat/completions?api-version=2024-02-15-preview",
  "CLAUDE_MEM_OPENAI_COMPATIBLE_MODEL": "gpt-4o"
}

Controls automatic generation of CLAUDE.md files in subdirectories.

{
  "CLAUDE_MEM_FOLDER_CLAUDEMD_ENABLED": "false"
}

• "false" (default): Disabled - no scattered CLAUDE.md files
• "true": Enabled - generates CLAUDE.md in modified folders

Fixed foreign key constraint errors for Gemini/OpenRouter/OpenAI-compatible agents.

Problem: Non-SDK agents failed to store observations because memory_session_id was NULL.

Solution: Set memory_session_id = contentSessionId at session start for non-SDK agents.

Same as original:

> /plugin marketplace add thedotmack/claude-mem
> /plugin install claude-mem

Then configure the OpenAI Compatible provider in settings.

This fork is designed for easy sync with upstream:

• All fork code marked with // === Fork: ... === or /* fork */
• Fork-specific configs at end of files (not middle)
• New file: src/services/worker/OpenAICompatibleAgent.ts

# Fetch upstream changes
git fetch upstream

# Merge upstream
git merge upstream/main

# Resolve conflicts (fork markers help locate changes)
grep -r "Fork:" src/ --include="*.ts"

# Verify build
bun build src/services/worker-service.ts --outdir=/tmp/test --target=bun

# Rebuild plugin
npm run build-and-sync

• Original Repository: https://github.com/thedotmack/claude-mem
• Original Documentation: https://docs.claude-mem.ai
• This Fork: https://github.com/geq1fan/claude-mem

Same as original: AGPL-3.0