feat: MCP server design spike — expose UnifiedFittingConfig via Model Context Protocol

[Anselmoo]

Summary

Design and implement a Model Context Protocol (MCP) server that exposes SpectraFit's fitting pipeline as LLM-callable tools.

Prerequisite

• Phases 1–5 complete — UnifiedFittingConfig is stable with JSON Schema via model_json_schema()
• FitResult(BaseModel) added in Phase 8 — stable, serialisable JSON output
• BatchFittingConfig(BaseModel) added in Phase 8 — multi-spectrum parallel fitting

Proposed Tools

Tool	Input	Output
fit_spectrum	UnifiedFittingConfig JSON	FitResult JSON
batch_fit	BatchFittingConfig JSON	list[FitResult] JSON
validate_config	Raw dict	Validation errors or schema
list_models	None	Available peak model names
get_schema	Model name	JSON Schema for that model

Technical Notes

• Use FastMCP or the official MCP Python SDK
• UnifiedFittingConfig.model_json_schema() is the stable API surface for input schema
• FitResult.model_json_schema() is the stable API surface for output schema
• Transport: stdio (default) + optional http/sse for remote access
• Authentication: API key via environment variable SPECTRAFIT_MCP_API_KEY

Acceptance Criteria

• MCP server starts with python -m spectrafit.mcp or spectrafit-mcp
• fit_spectrum tool validates input against UnifiedFittingConfig schema
• fit_spectrum tool returns FitResult as JSON
• batch_fit tool accepts BatchFittingConfig and returns list of results
• Server tested with Claude Desktop or another MCP client
• Documentation added to docs/mcp.md

References

• MCP Python SDK
• FastMCP
• spectrafit/models/fit_result.py — FitResult model
• spectrafit/core/fitting_config.py — UnifiedFittingConfig

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.