chatstrata 0.2.0__tar.gz → 0.2.2__tar.gz

{chatstrata-0.2.0 → chatstrata-0.2.2}/CHANGELOG.md

@@ -1,5 +1,13 @@
 # Changelog
 
+## 0.2.2 - 2026-07-29
+
+- Cap the `mcp` dependency at `>=1.12.0,<2`. `mcp` 2.0.0 removed the `mcp.server.fastmcp` module that `chatstrata/mcp/server.py` imports, so the previously unbounded pin resolved 2.0.0 and crashed `chatstrata-mcp` on startup with `ModuleNotFoundError: No module named 'mcp.server.fastmcp'`.
+
+## 0.2.1 - 2026-07-22
+
+- Fix `claude_code` conversation `project` paths: derive the project from the lossless `cwd` recorded in each transcript instead of decoding the session folder name. The folder-name encoding collapses `/`, `_`, `-`, and `.` all into `-`, so decoding it produced wrong, non-existent paths (e.g. `bstaq_git/pepstaq` became `bstaq/git`). New ingests are now correct automatically; `codex_cli` and `opencode` were unaffected.
+
 ## 0.2.0 - 2026-06-16
 
 - Add `chatstrata schedule` command group for automatic background sync.

{chatstrata-0.2.0 → chatstrata-0.2.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: chatstrata
-Version: 0.2.0
+Version: 0.2.2
 Summary: A personal, queryable archive of your AI conversations across providers
 Project-URL: Homepage, https://github.com/brandonbosch/chatstrata
 Project-URL: Repository, https://github.com/brandonbosch/chatstrata
@@ -34,7 +34,7 @@ Provides-Extra: embeddings
 Requires-Dist: sentence-transformers>=3.0; extra == 'embeddings'
 Requires-Dist: torch>=2.0; extra == 'embeddings'
 Provides-Extra: mcp
-Requires-Dist: mcp[cli]>=1.12.0; extra == 'mcp'
+Requires-Dist: mcp[cli]<2,>=1.12.0; extra == 'mcp'
 Provides-Extra: redact
 Requires-Dist: presidio-analyzer>=2.2; extra == 'redact'
 Requires-Dist: presidio-anonymizer>=2.2; extra == 'redact'
@@ -75,30 +75,64 @@ Concretely, with chatstrata you can:
 
 ## Status
 
-**Early alpha.** v0 includes adapters for Claude Code, claude.ai exports, Codex
-CLI, and OpenCode. The architecture is built so that adding more sources
-(ChatGPT exports, Cursor, etc.) is the work of one adapter — see
-[docs/adapter-guide.md](docs/adapter-guide.md).
+**Alpha.** Published on [PyPI](https://pypi.org/project/chatstrata/). Includes
+adapters for Claude Code, claude.ai exports, Codex CLI, and OpenCode, plus
+scheduled background sync so your archive stays current without manual runs. The
+architecture is built so that adding more sources (ChatGPT exports, Cursor, etc.)
+is the work of one adapter — see [docs/adapter-guide.md](docs/adapter-guide.md).
 
 ## Quickstart
 
-Requires Python 3.10+. DuckDB is installed as a Python dependency; you do not
-need to install a separate DuckDB server or CLI.
+Requires Python 3.10+. We recommend [uv](https://docs.astral.sh/uv/) — it's the
+fastest way to get started (no virtualenv wrangling, no dependency conflicts).
+
+### 1. Install
 
 ```bash
 uv tool install "chatstrata[embeddings]"
-# or: pipx install "chatstrata[embeddings]"
+```
 
-# Create the local DuckDB archive and show detected sources
-chatstrata init
+<details>
+<summary>Alternative: pipx</summary>
 
-# Auto-detect local app transcripts, ingest them, and generate embeddings
-chatstrata ingest --auto
+```bash
+pipx install "chatstrata[embeddings]"
+```
+</details>
 
-# See what's there
-chatstrata stats
+### 2. Initialize and ingest
+
+```bash
+chatstrata init                  # create the local DuckDB archive
+chatstrata ingest --auto         # auto-detect transcripts and generate embeddings
+```
+
+### 3. Set up background sync
+
+Never think about syncing again — chatstrata can keep your archive up to date
+automatically in the background:
 
-# Run a query
+```bash
+chatstrata schedule install              # sync every 15 minutes (default)
+chatstrata schedule install --interval 1h   # or pick your own interval
+chatstrata schedule status               # check that it's running
+```
+
+On macOS this installs a launchd agent that runs `chatstrata ingest --auto`
+on the configured interval. Logs go to `~/Library/Logs/chatstrata/`. To stop
+background sync:
+
+```bash
+chatstrata schedule uninstall
+```
+
+> **Linux:** systemd scheduling is not yet implemented. Use cron as a workaround:
+> `*/15 * * * * chatstrata ingest --auto`
+
+### 4. Query your archive
+
+```bash
+chatstrata stats
 chatstrata query "SELECT model, COUNT(*) FROM messages GROUP BY model"
 ```
 

{chatstrata-0.2.0 → chatstrata-0.2.2}/README.md

@@ -32,30 +32,64 @@ Concretely, with chatstrata you can:
 
 ## Status
 
-**Early alpha.** v0 includes adapters for Claude Code, claude.ai exports, Codex
-CLI, and OpenCode. The architecture is built so that adding more sources
-(ChatGPT exports, Cursor, etc.) is the work of one adapter — see
-[docs/adapter-guide.md](docs/adapter-guide.md).
+**Alpha.** Published on [PyPI](https://pypi.org/project/chatstrata/). Includes
+adapters for Claude Code, claude.ai exports, Codex CLI, and OpenCode, plus
+scheduled background sync so your archive stays current without manual runs. The
+architecture is built so that adding more sources (ChatGPT exports, Cursor, etc.)
+is the work of one adapter — see [docs/adapter-guide.md](docs/adapter-guide.md).
 
 ## Quickstart
 
-Requires Python 3.10+. DuckDB is installed as a Python dependency; you do not
-need to install a separate DuckDB server or CLI.
+Requires Python 3.10+. We recommend [uv](https://docs.astral.sh/uv/) — it's the
+fastest way to get started (no virtualenv wrangling, no dependency conflicts).
+
+### 1. Install
 
 ```bash
 uv tool install "chatstrata[embeddings]"
-# or: pipx install "chatstrata[embeddings]"
+```
 
-# Create the local DuckDB archive and show detected sources
-chatstrata init
+<details>
+<summary>Alternative: pipx</summary>
 
-# Auto-detect local app transcripts, ingest them, and generate embeddings
-chatstrata ingest --auto
+```bash
+pipx install "chatstrata[embeddings]"
+```
+</details>
 
-# See what's there
-chatstrata stats
+### 2. Initialize and ingest
+
+```bash
+chatstrata init                  # create the local DuckDB archive
+chatstrata ingest --auto         # auto-detect transcripts and generate embeddings
+```
+
+### 3. Set up background sync
+
+Never think about syncing again — chatstrata can keep your archive up to date
+automatically in the background:
 
-# Run a query
+```bash
+chatstrata schedule install              # sync every 15 minutes (default)
+chatstrata schedule install --interval 1h   # or pick your own interval
+chatstrata schedule status               # check that it's running
+```
+
+On macOS this installs a launchd agent that runs `chatstrata ingest --auto`
+on the configured interval. Logs go to `~/Library/Logs/chatstrata/`. To stop
+background sync:
+
+```bash
+chatstrata schedule uninstall
+```
+
+> **Linux:** systemd scheduling is not yet implemented. Use cron as a workaround:
+> `*/15 * * * * chatstrata ingest --auto`
+
+### 4. Query your archive
+
+```bash
+chatstrata stats
 chatstrata query "SELECT model, COUNT(*) FROM messages GROUP BY model"
 ```
 

{chatstrata-0.2.0 → chatstrata-0.2.2}/chatstrata/__init__.py

@@ -2,6 +2,6 @@
 
 from chatstrata.core.migrations import LATEST_VERSION as SCHEMA_VERSION
 
-__version__ = "0.2.0"
+__version__ = "0.2.1"
 
 __all__ = ["SCHEMA_VERSION", "__version__"]

{chatstrata-0.2.0 → chatstrata-0.2.2}/chatstrata/sources/claude_code/adapter.py

@@ -49,16 +49,39 @@ def _parse_timestamp(s: Any) -> datetime | None:
 
 
 def _decode_project_dir(dir_name: str) -> str:
-    """Claude Code encodes the project cwd by replacing '/' with '-'.
+    """Decode Claude Code's encoded project-folder name back into a path.
 
     e.g. '-Users-alice-code-myproj' -> '/Users/alice/code/myproj'
-    This is lossy but a reasonable best effort.
+
+    This decode is *lossy and irreversible*: Claude Code collapses '/', '_',
+    '-', and '.' in the cwd all into '-' when naming the folder, so a '-' in
+    the folder name could have been any of those characters. As a result many
+    distinct cwds map to the same decoded path (e.g. both '.../my_proj' and
+    '.../my/proj' decode identically). Prefer the lossless cwd recorded inside
+    the transcript (see `_project_from_events`); this decode is only a
+    last-resort fallback for transcripts that never recorded a cwd.
     """
     if dir_name.startswith("-"):
         return "/" + dir_name[1:].replace("-", "/")
     return dir_name
 
 
+def _project_from_events(events: list[dict[str, Any]]) -> str | None:
+    """Return the real working directory recorded in the transcript.
+
+    Every Claude Code event carries the session cwd, e.g.
+    ``"cwd": "/Users/alice/code/my_project"``. This is the lossless source of
+    truth for the project path -- unlike the encoded folder name, which cannot
+    be reversed (see `_decode_project_dir`). Use the first event that records a
+    non-empty cwd; the cwd is constant across a session.
+    """
+    for ev in events:
+        cwd = ev.get("cwd")
+        if isinstance(cwd, str) and cwd:
+            return cwd
+    return None
+
+
 def _role_from_event(event: dict[str, Any]) -> Role | None:
     t = event.get("type")
     if t == "user":
@@ -153,6 +176,8 @@ class ClaudeCodeAdapter:
             return
         for jsonl in sorted(root.glob("*/*.jsonl")):
             session_id = jsonl.stem
+            # Lossy fallback only: parse() prefers the lossless cwd from the
+            # transcript and uses this decoded folder name just as a backstop.
             project = _decode_project_dir(jsonl.parent.name)
             yield ConversationHandle(
                 source_native_id=session_id,
@@ -228,7 +253,12 @@ class ClaudeCodeAdapter:
                 )
             )
 
-        project = handle.metadata.get("project") if handle.metadata else None
+        # Prefer the lossless cwd recorded in the transcript over the lossy
+        # folder-name decode carried on the handle. Fall back to the decoded
+        # folder name only for transcripts that never recorded a cwd.
+        project = _project_from_events(events)
+        if project is None and handle.metadata:
+            project = handle.metadata.get("project")
         # If no summary, fall back to the first user text block as a title.
         if title is None:
             for m in messages:

{chatstrata-0.2.0 → chatstrata-0.2.2}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "chatstrata"
-version = "0.2.0"
+version = "0.2.2"
 description = "A personal, queryable archive of your AI conversations across providers"
 readme = "README.md"
 requires-python = ">=3.10"
@@ -45,7 +45,7 @@ redact = [
     "spacy>=3.7",
 ]
 mcp = [
-    "mcp[cli]>=1.12.0",
+    "mcp[cli]>=1.12.0,<2",
 ]
 dev = [
     "pytest>=8.0",