blitz-cli 0.8.0__tar.gz → 0.10.0__tar.gz

blitz_cli-0.10.0/.github/workflows/homebrew.yml

@@ -0,0 +1,104 @@
+name: Update Homebrew tap
+
+# On every release tag, point the sparepartslabs/homebrew-tap formula at the new
+# PyPI sdist. Runs alongside publish.yml (same v* trigger) and waits for the
+# sdist to appear on PyPI before reading its canonical URL + sha256.
+on:
+  push:
+    tags:
+      - "v*"
+
+jobs:
+  bump:
+    runs-on: ubuntu-latest
+    steps:
+      - name: Resolve version
+        id: v
+        run: echo "version=${GITHUB_REF_NAME#v}" >> "$GITHUB_OUTPUT"
+
+      - name: Fetch sdist url + sha256 from PyPI (wait for publish)
+        id: pypi
+        run: |
+          VERSION="${{ steps.v.outputs.version }}"
+          for i in $(seq 1 30); do
+            JSON=$(curl -fsSL "https://pypi.org/pypi/blitz-cli/${VERSION}/json" || true)
+            URL=$(printf '%s' "$JSON" | jq -r '.urls[]? | select(.packagetype=="sdist") | .url')
+            SHA=$(printf '%s' "$JSON" | jq -r '.urls[]? | select(.packagetype=="sdist") | .digests.sha256')
+            if [ -n "$URL" ] && [ "$URL" != "null" ]; then break; fi
+            echo "sdist for ${VERSION} not on PyPI yet, retrying ($i)..."
+            sleep 10
+          done
+          if [ -z "$URL" ] || [ "$URL" = "null" ]; then
+            echo "::error::blitz-cli ${VERSION} sdist not found on PyPI"; exit 1
+          fi
+          echo "url=$URL" >> "$GITHUB_OUTPUT"
+          echo "sha=$SHA" >> "$GITHUB_OUTPUT"
+
+      - name: Checkout tap
+        uses: actions/checkout@v4
+        with:
+          repository: sparepartslabs/homebrew-tap
+          token: ${{ secrets.HOMEBREW_TAP_TOKEN }}
+
+      - name: Update formula (url, sha256, and vendored resources)
+        env:
+          VERSION: ${{ steps.v.outputs.version }}
+          URL: ${{ steps.pypi.outputs.url }}
+          SHA: ${{ steps.pypi.outputs.sha }}
+        run: |
+          F=Formula/blitz-cli.rb
+          sed -i -E "s#^  url \".*\"#  url \"${URL}\"#" "$F"
+          sed -i -E "s#^  sha256 \".*\"#  sha256 \"${SHA}\"#" "$F"
+
+          # Refresh the vendored `resource` blocks for blitz-cli's runtime deps
+          # (rich + its transitive deps). `brew update-python-resources` can't be
+          # used on release day (it skips PyPI uploads < 1 day old), so resolve
+          # the deps in a throwaway venv and read each sdist url+sha256 from PyPI.
+          python3 - "$F" "$VERSION" <<'PY'
+          import json, os, re, subprocess, sys, tempfile, urllib.request, venv
+
+          formula, version = sys.argv[1], sys.argv[2]
+          tmp = tempfile.mkdtemp()
+          venv.create(f"{tmp}/v", with_pip=True)
+          pip = f"{tmp}/v/bin/pip"
+          subprocess.check_call([pip, "install", "-q", f"blitz-cli=={version}"])
+          freeze = subprocess.check_output([pip, "list", "--format=freeze"]).decode()
+
+          skip = {"blitz-cli", "pip", "setuptools", "wheel"}
+          deps = []
+          for line in freeze.splitlines():
+              if "==" in line:
+                  name, ver = line.split("==", 1)
+                  if name.lower() not in skip:
+                      deps.append((name, ver))
+
+          blocks = []
+          for name, ver in sorted(deps, key=lambda x: x[0].lower()):
+              d = json.load(urllib.request.urlopen(f"https://pypi.org/pypi/{name}/{ver}/json"))
+              sd = next((u for u in d["urls"] if u["packagetype"] == "sdist"), None)
+              if not sd:
+                  sys.exit(f"no sdist for {name} {ver}")
+              blocks.append(
+                  f'  resource "{d["info"]["name"]}" do\n'
+                  f'    url "{sd["url"]}"\n'
+                  f'    sha256 "{sd["digests"]["sha256"]}"\n'
+                  f'  end'
+              )
+
+          text = open(formula).read()
+          # drop existing resource blocks, then insert fresh ones after depends_on
+          text = re.sub(r'\n  resource "[^"]+" do\n(?:.*\n)*?  end\n', '\n', text)
+          inject = "\n" + "\n\n".join(blocks) + "\n"
+          text = re.sub(r'(  depends_on "python@[^"]+"\n)', r'\1' + inject, text, count=1)
+          text = re.sub(r'\n{3,}', '\n\n', text)  # collapse blank runs (idempotent)
+          open(formula, "w").write(text)
+          print(f"refreshed {len(blocks)} resources: {', '.join(n for n, _ in deps)}")
+          PY
+
+          git config user.name  "github-actions[bot]"
+          git config user.email "github-actions[bot]@users.noreply.github.com"
+          if git diff --quiet; then
+            echo "formula already up to date"; exit 0
+          fi
+          git commit -am "blitz-cli ${VERSION}"
+          git push

{blitz_cli-0.8.0 → blitz_cli-0.10.0}/.gitignore

@@ -8,4 +8,4 @@ dist/
 
 # stray test artifacts from running the CLI in-repo
 /.claude/
-/.playbooks/
+/.blitz/

blitz_cli-0.10.0/.mcp.json

@@ -0,0 +1,8 @@
+{
+  "mcpServers": {
+    "locker-room": {
+      "type": "sse",
+      "url": "http://localhost:8848/sse"
+    }
+  }
+}

{blitz_cli-0.8.0 → blitz_cli-0.10.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: blitz-cli
-Version: 0.8.0
+Version: 0.10.0
 Summary: Developer CLI for Blitz locker-room: spec-driven playbooks, a codebase/huddle ontology graph with an MCP tool, and the composable watcher/trainer/graph containers.
 License: MIT
 License-File: LICENSE
@@ -31,18 +31,18 @@ The core CLI is stdlib plus `rich`. Two optional extras carry heavier deps only
 when you need them: `blitz-cli[locker-room]` (the MCP server) and
 `blitz-cli[train]` (the QLoRA trainer).
 
-Top-level commands: `playbooks`, `locker-room`, `activity`.
+Top-level commands: `init`, `implement`, `locker-room`.
 
 ## Playbooks: spec-driven development commands
 
-`blitz playbooks init` installs a set of spec-driven-development slash-commands
+`blitz init` installs a set of spec-driven-development slash-commands
 into a repo, rendered for whichever agent you use (Claude Code, Cursor, GitHub
-Copilot, Gemini CLI, OpenCode), plus a shared `.playbooks/` working area.
+Copilot, Gemini CLI, OpenCode), plus a shared `.blitz/` working area.
 
 ```bash
-blitz playbooks init                 # auto-detect the agent(s) in this repo
-blitz playbooks init --agent claude  # or target one explicitly
-blitz playbooks init --all           # install for every supported agent
+blitz init                 # auto-detect the agent(s) in this repo
+blitz init --agent claude  # or target one explicitly
+blitz init --all           # install for every supported agent
 ```
 
 The commands are `specify`, `plan`, `tasks`, `implement`, `playbook`, `clarify`,
@@ -51,6 +51,28 @@ and `analyze`. Run at a folder of repos, `init` installs into each repo it finds
 for coordinating a feature across repos. Use `--ignore PATH` to skip a repo and
 `--root` to force a single install.
 
+## Implement: execute a tasks.md with a local model
+
+`blitz implement` drives a local model (served by Ollama) through a
+playbook-generated `tasks.md`, one task at a time. The model only proposes
+actions (full-file writes and shell commands); the driver applies them, runs
+each phase's machine-runnable verification, retries with the failure fed back
+(default 2), snapshots every file before overwriting it, and checks off
+`- [x]` in the tasks.md. A task that keeps failing stops the run so the state
+stays inspectable. Per-attempt model replies, snapshots, and exact token
+accounting (`usage.json`, per task and total) land under
+`.blitz/implement/runs/<timestamp>/`.
+
+```bash
+blitz implement                                   # auto-detect the open specs/*/tasks.md
+blitz implement specs/001-my-feature/tasks.md     # or name it
+blitz implement --model qwen2.5-14b-28k --start-at T005
+```
+
+Tasks whose text says "optional manual" are skipped and left unchecked. The
+Ollama endpoint comes from `--ollama-url` or `OLLAMA_HOST`; the model from
+`--model` or `BLITZ_IMPLEMENT_MODEL`.
+
 ## Locker-room: ontology graph, watcher, trainer, MCP
 
 The locker-room is a **context node**: a set of containers that ingest, build,
@@ -98,17 +120,17 @@ the base model with `BASE_MODEL_HF` or `blitz locker-room init --base-model`.
 
 ## Activity: replay a coding-agent session
 
-`blitz activity` reads your coding agent's own local session transcripts
+`blitz locker-room activity` reads your coding agent's own local session transcripts
 (currently Claude Code's, at `~/.claude/projects`) for the current repo and
 replays each prompt alongside the actions taken to resolve it (tool calls,
 sub-agents) and the final response:
 
 ```bash
-blitz activity --list               # sessions for this repo, most recent first
-blitz activity                      # the most recent session's timeline
-blitz activity <session-id-prefix>  # a specific session
-blitz activity --full               # untruncated prompts/responses/tool I/O
-blitz activity --html report.html   # also write a self-contained HTML report
+blitz locker-room activity --list               # sessions for this repo, most recent first
+blitz locker-room activity                      # the most recent session's timeline
+blitz locker-room activity <session-id-prefix>  # a specific session
+blitz locker-room activity --full               # untruncated prompts/responses/tool I/O
+blitz locker-room activity --html report.html   # also write a self-contained HTML report
 ```
 
 The terminal view is compact by default; `--full` shows untruncated text. The

{blitz_cli-0.8.0 → blitz_cli-0.10.0}/README.md

@@ -12,18 +12,18 @@ The core CLI is stdlib plus `rich`. Two optional extras carry heavier deps only
 when you need them: `blitz-cli[locker-room]` (the MCP server) and
 `blitz-cli[train]` (the QLoRA trainer).
 
-Top-level commands: `playbooks`, `locker-room`, `activity`.
+Top-level commands: `init`, `implement`, `locker-room`.
 
 ## Playbooks: spec-driven development commands
 
-`blitz playbooks init` installs a set of spec-driven-development slash-commands
+`blitz init` installs a set of spec-driven-development slash-commands
 into a repo, rendered for whichever agent you use (Claude Code, Cursor, GitHub
-Copilot, Gemini CLI, OpenCode), plus a shared `.playbooks/` working area.
+Copilot, Gemini CLI, OpenCode), plus a shared `.blitz/` working area.
 
 ```bash
-blitz playbooks init                 # auto-detect the agent(s) in this repo
-blitz playbooks init --agent claude  # or target one explicitly
-blitz playbooks init --all           # install for every supported agent
+blitz init                 # auto-detect the agent(s) in this repo
+blitz init --agent claude  # or target one explicitly
+blitz init --all           # install for every supported agent
 ```
 
 The commands are `specify`, `plan`, `tasks`, `implement`, `playbook`, `clarify`,
@@ -32,6 +32,28 @@ and `analyze`. Run at a folder of repos, `init` installs into each repo it finds
 for coordinating a feature across repos. Use `--ignore PATH` to skip a repo and
 `--root` to force a single install.
 
+## Implement: execute a tasks.md with a local model
+
+`blitz implement` drives a local model (served by Ollama) through a
+playbook-generated `tasks.md`, one task at a time. The model only proposes
+actions (full-file writes and shell commands); the driver applies them, runs
+each phase's machine-runnable verification, retries with the failure fed back
+(default 2), snapshots every file before overwriting it, and checks off
+`- [x]` in the tasks.md. A task that keeps failing stops the run so the state
+stays inspectable. Per-attempt model replies, snapshots, and exact token
+accounting (`usage.json`, per task and total) land under
+`.blitz/implement/runs/<timestamp>/`.
+
+```bash
+blitz implement                                   # auto-detect the open specs/*/tasks.md
+blitz implement specs/001-my-feature/tasks.md     # or name it
+blitz implement --model qwen2.5-14b-28k --start-at T005
+```
+
+Tasks whose text says "optional manual" are skipped and left unchecked. The
+Ollama endpoint comes from `--ollama-url` or `OLLAMA_HOST`; the model from
+`--model` or `BLITZ_IMPLEMENT_MODEL`.
+
 ## Locker-room: ontology graph, watcher, trainer, MCP
 
 The locker-room is a **context node**: a set of containers that ingest, build,
@@ -79,17 +101,17 @@ the base model with `BASE_MODEL_HF` or `blitz locker-room init --base-model`.
 
 ## Activity: replay a coding-agent session
 
-`blitz activity` reads your coding agent's own local session transcripts
+`blitz locker-room activity` reads your coding agent's own local session transcripts
 (currently Claude Code's, at `~/.claude/projects`) for the current repo and
 replays each prompt alongside the actions taken to resolve it (tool calls,
 sub-agents) and the final response:
 
 ```bash
-blitz activity --list               # sessions for this repo, most recent first
-blitz activity                      # the most recent session's timeline
-blitz activity <session-id-prefix>  # a specific session
-blitz activity --full               # untruncated prompts/responses/tool I/O
-blitz activity --html report.html   # also write a self-contained HTML report
+blitz locker-room activity --list               # sessions for this repo, most recent first
+blitz locker-room activity                      # the most recent session's timeline
+blitz locker-room activity <session-id-prefix>  # a specific session
+blitz locker-room activity --full               # untruncated prompts/responses/tool I/O
+blitz locker-room activity --html report.html   # also write a self-contained HTML report
 ```
 
 The terminal view is compact by default; `--full` shows untruncated text. The

{blitz_cli-0.8.0 → blitz_cli-0.10.0}/blitz_cli/__init__.py

@@ -13,7 +13,7 @@ Commands:
 - ``blitz traces list`` / ``blitz traces show <trace_id>`` — inspect captured
   traces from the local SQLite store.
 
-- ``blitz activity [session] [--list] [--html report.html]`` — replay a
+- ``blitz locker-room activity [session] [--list] [--html report.html]`` — replay a
   Claude Code session for this repo: each prompt you sent, the actions taken
   to resolve it (tool calls, sub-agents), and the final response. Reads
   ``~/.claude/projects`` directly; unrelated to Blitz's own captured traces.
@@ -30,4 +30,4 @@ Commands:
 
 __all__ = ["__version__"]
 
-__version__ = "0.6.0"
+__version__ = "0.10.0"

{blitz_cli-0.8.0 → blitz_cli-0.10.0}/blitz_cli/_activity.py

@@ -1,4 +1,4 @@
-"""`blitz activity` — replay a coding-agent session: what you asked, and how it
+"""`blitz locker-room activity` — replay a coding-agent session: what you asked, and how it
 was resolved.
 
 The currently-supported transcript format is Claude Code's, at
@@ -82,6 +82,83 @@ def default_sessions_dir() -> Path:
     return Path.home() / ".claude" / "projects"
 
 
+def _zero_usage(**extra) -> dict:
+    return {"input": 0, "output": 0, "cache_read": 0, "cache_creation": 0, **extra}
+
+
+def _add_usage(agg: dict, usage: dict) -> None:
+    agg["input"] += usage.get("input_tokens") or 0
+    agg["output"] += usage.get("output_tokens") or 0
+    agg["cache_read"] += usage.get("cache_read_input_tokens") or 0
+    agg["cache_creation"] += usage.get("cache_creation_input_tokens") or 0
+
+
+def frontier_usage_by_slug(sessions_root: Path, slugs) -> dict:
+    """Attribute frontier-agent (Claude Code) token spend to feature specs by
+    scanning session transcripts under ``sessions_root`` for ``specs/<slug>``
+    mentions, at two granularities:
+
+    - ``session``: the whole session's usage goes to the single most-mentioned
+      slug (sessions mentioning no spec are ignored). An upper bound — long
+      sessions that merely discuss a spec inflate it.
+    - ``turn``: each assistant message is attributed to the slug most recently
+      mentioned at that point in the transcript, so usage before the spec
+      comes up — or after the session moves on to another spec — is excluded.
+      The closer proxy for implementation cost.
+
+    Usage is deduplicated by API message id (a transcript writes several
+    entries per message, each repeating the same usage block). Returns
+    ``{slug: {"session": {input, output, cache_read, cache_creation, sessions},
+              "turn": {input, output, cache_read, cache_creation, messages}}}``."""
+    needles = {slug: f"specs/{slug}" for slug in slugs}
+    totals: dict = {}
+
+    def slot(slug: str) -> dict:
+        return totals.setdefault(slug, {
+            "session": _zero_usage(sessions=0),
+            "turn": _zero_usage(messages=0),
+        })
+
+    for jsonl in sorted(sessions_root.glob("*/*.jsonl")):
+        try:
+            text = jsonl.read_text(encoding="utf-8", errors="replace")
+        except OSError:
+            continue
+        counts = {slug: text.count(n) for slug, n in needles.items()}
+        best = max(counts, key=counts.get) if counts else None
+        if not best or counts[best] == 0:
+            continue
+        current: str | None = None  # sticky most-recently-mentioned slug
+        by_message: dict = {}       # message id -> (usage, slug at that point)
+        for i, line in enumerate(text.splitlines()):
+            mentioned = [(line.find(n), slug) for slug, n in needles.items() if n in line]
+            if mentioned:
+                current = max(mentioned)[1]  # last mention in the line wins
+            if '"usage"' not in line:
+                continue
+            try:
+                rec = json.loads(line)
+            except ValueError:
+                continue
+            if rec.get("type") != "assistant":
+                continue
+            message = rec.get("message", {})
+            usage = message.get("usage")
+            if usage:
+                by_message[message.get("id") or f"line:{i}"] = (usage, current)
+        if not by_message:
+            continue
+        session_agg = slot(best)["session"]
+        for usage, slug in by_message.values():
+            _add_usage(session_agg, usage)
+            if slug is not None:
+                turn_agg = slot(slug)["turn"]
+                _add_usage(turn_agg, usage)
+                turn_agg["messages"] += 1
+        session_agg["sessions"] += 1
+    return totals
+
+
 def default_file_history_dir() -> Path:
     """Root of Claude Code's own pre-edit file backups, one subdir per
     session id — a separate store from ~/.claude/projects, linked to it via
@@ -657,7 +734,7 @@ def render_html(session: Session, limit: int) -> str:
         '<html lang="en">',
         "<head>",
         '<meta charset="utf-8">',
-        f"<title>blitz activity — {html.escape(session.title)}</title>",
+        f"<title>blitz locker-room activity — {html.escape(session.title)}</title>",
         f"<style>{_HTML_STYLE}</style>",
         "</head>",
         "<body>",