Emil_Shanaty/memwalk
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

v0.4.3: agent-friendly MCP tool descriptions with step-by-step workflow

Browse Source
This commit is contained in:
emil
2026-05-16 18:47:09 +03:00
parent 974aba45eb
commit 3d0e087ebd
3 changed files with 34 additions and 24 deletions
Download Patch File Download Diff File Expand all files Collapse all files
memwalk/__init__.py
+1 -1
View File
@@ -1,3 +1,3 @@
"""memwalk — ask AI about any codebase via cached SSM state."""
__version__ = "0.4.2"
__version__ = "0.4.3"
memwalk/mcp_server.py
+32 -22
View File
@@ -40,10 +40,11 @@ async def list_tools() -> list[Tool]:
description=(
"Read all source files under the given directory and build a "
"cached SSM state that can be queried in subsequent ask() calls. "
"Slow first time (5-30s for medium repos, longer for big ones); "
"Use this for small-to-medium repos that fit in a single context "
"window (default ~120K chars). For large repos that exceed the "
"context limit, use digest_split instead. Slow first time (5-30s); "
"cache is reused on subsequent calls until source files change. "
"Use before ask() to control when ingestion happens, or just call "
"ask() directly which will auto-digest as needed."
"You can also call ask() directly which auto-digests as needed."
),
inputSchema={
"type": "object",
@@ -68,13 +69,15 @@ async def list_tools() -> list[Tool]:
Tool(
name="ask",
description=(
"Query a codebase using its cached SSM state. Returns the "
"model's answer based on the previously digested source. "
"Auto-digests if no fresh cache exists (first call may be "
"slow). Subsequent calls on the same codebase are fast "
"(<1s typical). Best for descriptive questions: 'what does "
"module X do', 'where is concept Y used', 'list all CLI "
"commands', 'how would I add feature Z'."
"Step 3 in the multi-corpus workflow. Query a codebase using its "
"cached SSM state. Returns the model's answer based on the previously "
"digested source. For large repos that were split-digested, you must "
"pass the specific subdirectory path in 'path' that is most relevant "
"to the question — not the repo root. For example: "
"ask('/repo/src/auth', 'how does login work?') or "
"ask('/repo/backend/db', 'what migrations exist?'). "
"Auto-digests if no fresh cache exists (first call may be slow). "
"Subsequent calls on the same codebase are fast (<1s typical)."
),
inputSchema={
"type": "object",
@@ -117,16 +120,19 @@ async def list_tools() -> list[Tool]:
Tool(
name="list_subdirs",
description=(
"List subdirectories of a codebase root with file counts, "
"estimated char sizes, depth, and cache status. Use this before "
"digest_split to see which subdirectories are available. "
"Cheap — does not load the model."
"Step 1 in the multi-corpus workflow. Lists subdirectories of a "
"codebase root with file counts, estimated char sizes, recursion "
"depth, and cache status. For large repos, this shows you what leaf "
"directories are available so you can pick the right ones to digest. "
"Directories that are too large for the current n_ctx budget are "
"shown with their depth; you can use max_depth to control how deep "
"the recursion goes. Cheap — does not load the model."
),
inputSchema={
"type": "object",
"properties": {
"path": {"type": "string", "description": "Codebase root to inspect."},
"max_depth": {"type": "integer", "description": "Max recursion depth (default: unlimited)."},
"max_depth": {"type": "integer", "description": "Max recursion depth (default: unlimited). Use 1 for immediate children only."},
},
"required": ["path"],
},
@@ -134,19 +140,23 @@ async def list_tools() -> list[Tool]:
Tool(
name="digest_split",
description=(
"Discover subdirectories under the given path and digest each "
"independently into its own cached SSM state. Each subdirectory gets "
"a separate cache, so subsequent ask() calls can target specific "
"sub-caches. Use list_subdirs first to see what will be digested. "
"Slow first time; cache is reused on subsequent calls."
"Step 2 in the multi-corpus workflow. Discovers subdirectories under "
"the given path and digests each independently into its own cached "
"SSM state. Each subdirectory gets a separate cache keyed by its "
"absolute path, so subsequent ask() calls can target specific "
"sub-caches (e.g. ask '/repo/src/auth' 'how does login work?'). "
"Use list_subdirs first to see what will be digested. Slow first "
"time (5-30s per subdirectory); cache is reused on subsequent calls. "
"Large subdirectories that exceed the n_ctx budget are skipped with "
"an error unless you increase max_depth to split them further."
),
inputSchema={
"type": "object",
"properties": {
"path": {"type": "string", "description": "Codebase root to split-digest."},
"force": {"type": "boolean", "description": "Re-ingest even if cache is fresh.", "default": False},
"n_ctx": {"type": "integer", "description": "Override config n_ctx for this digest."},
"max_depth": {"type": "integer", "description": "Max recursion depth (default: unlimited)."},
"n_ctx": {"type": "integer", "description": "Override config n_ctx for this digest (default: auto-detected from free VRAM)."},
"max_depth": {"type": "integer", "description": "Max recursion depth (default: unlimited). Use 1 for immediate children only."},
},
"required": ["path"],
},
pyproject.toml
+1 -1
View File
@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
[project]
name = "memwalk"
version = "0.4.2"
version = "0.4.3"
description = "Ask AI about any codebase — local, cached, SSM-state-backed exploration via memba + Nemotron"
readme = "README.md"
license = { text = "MIT" }