Phase 4-A: Backend CLAUDE.md Memory Browser #50

New Issue

Open

opened 2026-05-24 16:55:19 -05:00 by null · 0 comments

null commented

2026-05-24 16:55:19 -05:00

Owner

Source plan: /home/kaspa/.claude/plans/with-our-backend-created-precious-owl.md

Feature context: Feature 4: CLAUDE.md Memory Browser

Scope

Phase 4-A — Backend

Summary

Add GET /claude-code/projects/{encoded_project_dir}/memory that walks the project's working directory for CLAUDE.md files and returns their content.

Problem

Users lose track of what instructions are in their CLAUDE.md files across projects. Pipeline knows each project's working directory from session data but does not expose the memory files.

Affected area

Backend service: Claude Code reader
API: /claude-code

Affected files

backend/app/services/claude_code_reader.py — add get_project_memory_files() function
backend/app/api/claude_code.py — add new route
backend/app/schemas/claude_code.py — add MemoryFileRead, ProjectMemoryResponse schemas

Affected routes or endpoints

GET /api/v1/claude-code/projects/{project_dir}/memory
- project_dir is URL-encoded project directory name (the ~/.claude/projects/ subdir name)

Expected behavior

Reads the cwd for the project from session data
Walks the cwd directory tree for files named CLAUDE.md (case-insensitive), up to 3 levels deep
Returns list of {relative_path, content, size_bytes, last_modified} objects
Also reads the global ~/.claude/CLAUDE.md if it exists and includes it as {relative_path: "~/.claude/CLAUDE.md", ...}
Returns empty list if project cwd is unknown or no CLAUDE.md files exist
File content truncated at 50 000 chars

Steps to reproduce (acceptance criteria)

Call GET /api/v1/claude-code/projects/{any_project}/memory
Returns at least the global ~/.claude/CLAUDE.md if it exists
Returns project-level CLAUDE.md files if present
Returns empty array (not 404) when no memory files found

Source plan: `/home/kaspa/.claude/plans/with-our-backend-created-precious-owl.md` Feature context: **Feature 4: CLAUDE.md Memory Browser** ## Scope ### Phase 4-A — Backend #### Summary Add `GET /claude-code/projects/{encoded_project_dir}/memory` that walks the project's working directory for `CLAUDE.md` files and returns their content. #### Problem Users lose track of what instructions are in their CLAUDE.md files across projects. Pipeline knows each project's working directory from session data but does not expose the memory files. #### Affected area - Backend service: Claude Code reader - API: `/claude-code` #### Affected files - `backend/app/services/claude_code_reader.py` — add `get_project_memory_files()` function - `backend/app/api/claude_code.py` — add new route - `backend/app/schemas/claude_code.py` — add `MemoryFileRead`, `ProjectMemoryResponse` schemas #### Affected routes or endpoints - `GET /api/v1/claude-code/projects/{project_dir}/memory` - `project_dir` is URL-encoded project directory name (the `~/.claude/projects/` subdir name) #### Expected behavior - Reads the `cwd` for the project from session data - Walks the `cwd` directory tree for files named `CLAUDE.md` (case-insensitive), up to 3 levels deep - Returns list of `{relative_path, content, size_bytes, last_modified}` objects - Also reads the global `~/.claude/CLAUDE.md` if it exists and includes it as `{relative_path: "~/.claude/CLAUDE.md", ...}` - Returns empty list if project `cwd` is unknown or no CLAUDE.md files exist - File content truncated at 50 000 chars #### Steps to reproduce (acceptance criteria) 1. Call `GET /api/v1/claude-code/projects/{any_project}/memory` 2. Returns at least the global `~/.claude/CLAUDE.md` if it exists 3. Returns project-level CLAUDE.md files if present 4. Returns empty array (not 404) when no memory files found ---