coulomb/agentic-resources
2
0
Fork 0
generated from coulomb/repo-seed
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

session-memory: weekly retro entrypoint + hub publish (AGENTIC-WP-0010)

Browse Source 
The analysis half of the weekly coding retrospection. retro/build.py: windowed
detect+measure -> top-3 improvement suggestions per repo (cross-flavor first,
recommendations pulled from the Pattern Catalog) + fleet snapshot. retro/publish.py:
publishes the report to the hub as the coding_retro read model (event_type=
coding_retro progress event) + local JSON/md, graceful degrade. retro entrypoint
with --window-days/--publish/--json. Live verify over real sessions surfaced
per-repo suggestions with catalog recommendations. 13 new tests; suite 152/152.

Consumed by activity-core ACTIVITY-WP-0008 (Weekly Coding Retrospection, Sat 19:00).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This commit is contained in:
Bernd Worsch
2026-06-07 19:17:24 +02:00
parent 15ba625351
commit 0d05dfcc5d
12 changed files with 932 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

21
session_memory/README.md
View File

@@ -42,6 +42,9 @@ session_memory/
measure/metrics.py # fleet metrics + persisted baseline snapshots
measure/effect.py # before/after per-pattern effectiveness
measure/__main__.py # python -m session_memory.measure
retro/build.py # windowed top-3-per-repo suggestions
retro/publish.py # hub coding_retro read model + local report
retro/__main__.py # python -m session_memory.retro
config.toml # store paths, retention caps, sources, repo->domain map, curate gate
```
@@ -163,6 +166,24 @@ python -m session_memory.measure --no-save --json
retired. Recorded pre-fix baseline (2026-06-07): 27 sessions, infra-overhead
median 11.7 %, error rate 0.96, schema-thrash 8 sessions.
## Weekly retro (the input to the scheduled retrospection)
A windowed roll-up: detect + measure over the last N days → the **top-3
improvement suggestions per repo** (cross-flavor first; recommendations pulled
from the Pattern Catalog) → published to the hub as the `coding_retro` read model.
```bash
python -m session_memory.retro # last 7 days, local report
python -m session_memory.retro --window-days 30 --json
python -m session_memory.retro --publish # also post coding_retro to the hub
```
Writes `retro/last_retro.{json,md}` and (with `--publish`) posts an
`event_type=coding_retro` progress event. This is consumed by activity-core's
**Weekly Coding Retrospection** schedule (ACTIVITY-WP-0008, Saturday 19:00 Berlin),
which emits one improvement task per relevant repo. Hub publish degrades
gracefully when the hub is unreachable.
## Retention knobs (`[retention]` in config.toml)
| Key | Meaning |

8
session_memory/config.toml
View File

@@ -43,6 +43,14 @@ min_prompt_len = 25 # first prompt shorter than this is treated as trivial
[measure]
baselines = "session_memory/measure/baselines.jsonl" # timestamped metric snapshots (committed)
# Weekly retro (AGENTIC-WP-0010): windowed top-3-per-repo report, published to the
# hub as the coding_retro read model that activity-core's weekly schedule consumes.
[retro]
window_days = 7
report_json = "session_memory/retro/last_retro.json" # latest report (committed)
report_md = "session_memory/retro/last_retro.md" # human-readable mirror
hub_url = "http://127.0.0.1:8000" # for --publish (best-effort)
# Distribute phase (AGENTIC-WP-0007): where per-flavor proposals + the active
# registry are written. Proposals are HITL — reviewed, never auto-applied.
[distribute]

9
session_memory/retro/__init__.py Normal file
View File

@@ -0,0 +1,9 @@
"""Weekly retro (AGENTIC-WP-0010) — the analysis half of the coding retrospection.
build.py windowed detect + measure -> ranked top-3 suggestions per repo (T01)
publish.py publish the retro to the hub read model + local report (T02)
__main__.py python -m session_memory.retro (T03)
Consumed by activity-core's weekly-coding-retro schedule (ACTIVITY-WP-0008) via
the ``event_type=coding_retro`` read model.
"""

68
session_memory/retro/__main__.py Normal file
View File

@@ -0,0 +1,68 @@
"""Weekly retro entrypoint (AGENTIC-WP-0010 T03).
python -m session_memory.retro [--window-days 7] [--since D] [--until D]
[--publish] [--json]
Builds the windowed top-3-per-repo retro over the captured sessions, writes a local
JSON + markdown report, and (with ``--publish``) posts it to the hub as the
``coding_retro`` read model that activity-core's weekly schedule consumes.
"""
from __future__ import annotations
import argparse
import json
import os
from ..core.store import Store
from ..curate.catalog import Catalog
from ..ingest import _expand, load_config
from .build import weekly_retro
from .publish import publish_to_hub, render_markdown, write_local
def run_retro(config: dict, *, window_days=None, since=None, until=None):
s = config.get("store", {})
store = Store(_expand(s["db_path"]), _expand(s["blob_dir"]))
digests = store.list_digests()
store.close()
cur = config.get("curate", {})
catalog = Catalog(_expand(cur.get("catalog_dir", "session_memory/catalog")))
rcfg = config.get("retro", {})
return weekly_retro(digests, catalog, since=since, until=until,
window_days=window_days or rcfg.get("window_days", 7))
def main(argv=None) -> int:
here = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
ap = argparse.ArgumentParser(description="Build (and optionally publish) the weekly coding retro.")
ap.add_argument("--config", default=os.path.join(here, "config.toml"))
ap.add_argument("--window-days", type=int, default=None)
ap.add_argument("--since", default=None)
ap.add_argument("--until", default=None)
ap.add_argument("--publish", action="store_true", help="post to the hub coding_retro read model")
ap.add_argument("--json", action="store_true")
args = ap.parse_args(argv)
config = load_config(args.config)
report = run_retro(config, window_days=args.window_days, since=args.since, until=args.until)
rcfg = config.get("retro", {})
write_local(report, _expand(rcfg.get("report_json", "session_memory/retro/last_retro.json")),
_expand(rcfg.get("report_md", "session_memory/retro/last_retro.md")))
published = None
if args.publish:
published = publish_to_hub(report, base_url=rcfg.get("hub_url", "http://127.0.0.1:8000"))
if args.json:
print(json.dumps({"report": report, "published": published}, indent=2))
else:
print(render_markdown(report))
if args.publish:
print(f"\npublished to hub: {published}")
return 0
if __name__ == "__main__":
raise SystemExit(main())

100
session_memory/retro/build.py Normal file
View File

@@ -0,0 +1,100 @@
"""Windowed weekly retro report (AGENTIC-WP-0010 T01).
Runs the existing detect pipeline over a date window, ranks the recurring problem
patterns into **per-repo improvement suggestions** (top 3, cross-flavor first),
attaches a recommendation from the Pattern Catalog where one exists, and bundles a
fleet measure snapshot for context. Pure function over digests — the entrypoint
(T03) handles store/publish.
"""
from __future__ import annotations
import collections
from dataclasses import asdict, dataclass
from datetime import datetime, timedelta, timezone
from typing import Optional
from ..curate.schema import SolutionPattern
from ..detect.cluster import cluster
from ..detect.quality import QualityConfig, filter_real
from ..detect.signals import extract_signals
from ..measure.metrics import aggregate
# score at/above which a suggestion is "high" priority even when single-flavor
_HIGH_SCORE = 100.0
def _parse(ts: str) -> datetime:
return datetime.fromisoformat(ts.replace("Z", "+00:00"))
def _iso(dt: datetime) -> str:
return dt.astimezone(timezone.utc).strftime("%Y-%m-%dT%H:%M:%SZ")
def _now() -> datetime:
return datetime.now(timezone.utc)
@dataclass
class Suggestion:
repo: str
title: str
recommendation: str
priority: str # high | medium
score: float
signal_type: str
cross_flavor: bool
pattern_key: str
def _recommendation(pattern_key: str, catalog) -> Optional[str]:
if catalog is None:
return None
sp = catalog.load(SolutionPattern.make_id(pattern_key))
if sp and sp.resolutions:
return sp.resolutions[0].summary
return None
def weekly_retro(digests: list[dict], catalog=None, *, since: Optional[str] = None,
until: Optional[str] = None, window_days: int = 7,
max_per_repo: int = 3, min_frequency: int = 2,
quality: Optional[QualityConfig] = None) -> dict:
"""Build the ranked weekly retro report over a date window."""
until_dt = _parse(until) if until else _now()
since_dt = _parse(since) if since else until_dt - timedelta(days=window_days)
windowed = [d for d in digests
if d.get("started_at") and since_dt <= _parse(d["started_at"]) < until_dt]
real = filter_real(windowed, quality or QualityConfig())
patterns = cluster(extract_signals(real), min_frequency=min_frequency)
by_repo: dict[str, list[Suggestion]] = collections.defaultdict(list)
for p in patterns:
if p.polarity != "problem":
continue # improvements come from problems
rec = (_recommendation(p.key, catalog)
or f"Investigate {p.signal_type.replace('_', ' ')} on {p.locus}")
priority = "high" if (p.cross_flavor or p.score >= _HIGH_SCORE) else "medium"
for repo in (p.repos or ["(unknown)"]):
by_repo[repo].append(Suggestion(
repo=repo, title=p.title, recommendation=rec, priority=priority,
score=p.score, signal_type=p.signal_type, cross_flavor=p.cross_flavor,
pattern_key=p.key))
suggestions: list[Suggestion] = []
for repo in sorted(by_repo):
items = sorted(by_repo[repo], key=lambda s: -s.score)
suggestions.extend(items[:max_per_repo])
# cross-flavor first, then by score (global ordering for the report)
suggestions.sort(key=lambda s: (not s.cross_flavor, -s.score))
return {
"window": {"since": _iso(since_dt), "until": _iso(until_dt), "days": window_days},
"generated_at": _iso(_now()),
"n_sessions": len(real),
"suggestions": [asdict(s) for s in suggestions],
"measure": aggregate(real),
}

322
session_memory/retro/last_retro.json Normal file
View File

@@ -0,0 +1,322 @@
{
"generated_at": "2026-06-07T17:14:00Z",
"measure": {
"error_rate": 0.957,
"infra_overhead_share_median": 0.167,
"infra_overhead_share_p90": 0.23,
"n_sessions": 23,
"recurring_error_occurrences": 463,
"schema_thrash_sessions": 7,
"success_rate": 1.0,
"tokens_p50": 250725,
"tokens_p90": 901422
},
"n_sessions": 23,
"suggestions": [
{
"cross_flavor": true,
"pattern_key": "problem:recurring_error:make: *** [makefile:<n>: fix-consistency] error <n>",
"priority": "high",
"recommendation": "Investigate recurring error on make: *** [makefile:<n>: fix-consistency] error <n>",
"repo": "net-kingdom",
"score": 54.0,
"signal_type": "recurring_error",
"title": "cross-flavor problem: recurring error"
},
{
"cross_flavor": false,
"pattern_key": "problem:tool_thrash:tool:Bash",
"priority": "high",
"recommendation": "Batch related shell work into one script, not many small Bash calls",
"repo": "activity-core",
"score": 13128.0,
"signal_type": "tool_thrash",
"title": "problem: tool thrash"
},
{
"cross_flavor": false,
"pattern_key": "problem:tool_thrash:tool:Bash",
"priority": "high",
"recommendation": "Batch related shell work into one script, not many small Bash calls",
"repo": "artifact-store",
"score": 13128.0,
"signal_type": "tool_thrash",
"title": "problem: tool thrash"
},
{
"cross_flavor": false,
"pattern_key": "problem:tool_thrash:tool:Bash",
"priority": "high",
"recommendation": "Batch related shell work into one script, not many small Bash calls",
"repo": "citation-evidence",
"score": 13128.0,
"signal_type": "tool_thrash",
"title": "problem: tool thrash"
},
{
"cross_flavor": false,
"pattern_key": "problem:tool_thrash:tool:Bash",
"priority": "high",
"recommendation": "Batch related shell work into one script, not many small Bash calls",
"repo": "infospace-bench",
"score": 13128.0,
"signal_type": "tool_thrash",
"title": "problem: tool thrash"
},
{
"cross_flavor": false,
"pattern_key": "problem:tool_thrash:tool:Bash",
"priority": "high",
"recommendation": "Batch related shell work into one script, not many small Bash calls",
"repo": "railiance-apps",
"score": 13128.0,
"signal_type": "tool_thrash",
"title": "problem: tool thrash"
},
{
"cross_flavor": false,
"pattern_key": "problem:tool_thrash:tool:Bash",
"priority": "high",
"recommendation": "Batch related shell work into one script, not many small Bash calls",
"repo": "state-hub",
"score": 13128.0,
"signal_type": "tool_thrash",
"title": "problem: tool thrash"
},
{
"cross_flavor": false,
"pattern_key": "problem:schema_thrash:schema_load",
"priority": "high",
"recommendation": "Load the tool schemas you'll need once, up front",
"repo": "activity-core",
"score": 441.0,
"signal_type": "schema_thrash",
"title": "problem: schema thrash"
},
{
"cross_flavor": false,
"pattern_key": "problem:schema_thrash:schema_load",
"priority": "high",
"recommendation": "Load the tool schemas you'll need once, up front",
"repo": "citation-evidence",
"score": 441.0,
"signal_type": "schema_thrash",
"title": "problem: schema thrash"
},
{
"cross_flavor": false,
"pattern_key": "problem:schema_thrash:schema_load",
"priority": "high",
"recommendation": "Load the tool schemas you'll need once, up front",
"repo": "flex-auth",
"score": 441.0,
"signal_type": "schema_thrash",
"title": "problem: schema thrash"
},
{
"cross_flavor": false,
"pattern_key": "problem:schema_thrash:schema_load",
"priority": "high",
"recommendation": "Load the tool schemas you'll need once, up front",
"repo": "infospace-bench",
"score": 441.0,
"signal_type": "schema_thrash",
"title": "problem: schema thrash"
},
{
"cross_flavor": false,
"pattern_key": "problem:schema_thrash:schema_load",
"priority": "high",
"recommendation": "Load the tool schemas you'll need once, up front",
"repo": "ops-bridge",
"score": 441.0,
"signal_type": "schema_thrash",
"title": "problem: schema thrash"
},
{
"cross_flavor": false,
"pattern_key": "problem:recurring_error:<tool_use_error>file has not been read yet. read it first before writing to it.<<path>>",
"priority": "high",
"recommendation": "Investigate recurring error on <tool_use_error>file has not been read yet. read it first before writing to it.<<path>>",
"repo": "activity-core",
"score": 290.0,
"signal_type": "recurring_error",
"title": "problem: recurring error"
},
{
"cross_flavor": false,
"pattern_key": "problem:recurring_error:<tool_use_error>file has not been read yet. read it first before writing to it.<<path>>",
"priority": "high",
"recommendation": "Investigate recurring error on <tool_use_error>file has not been read yet. read it first before writing to it.<<path>>",
"repo": "citation-evidence",
"score": 290.0,
"signal_type": "recurring_error",
"title": "problem: recurring error"
},
{
"cross_flavor": false,
"pattern_key": "problem:recurring_error:<tool_use_error>file has not been read yet. read it first before writing to it.<<path>>",
"priority": "high",
"recommendation": "Investigate recurring error on <tool_use_error>file has not been read yet. read it first before writing to it.<<path>>",
"repo": "infospace-bench",
"score": 290.0,
"signal_type": "recurring_error",
"title": "problem: recurring error"
},
{
"cross_flavor": false,
"pattern_key": "problem:recurring_error:<tool_use_error>file has not been read yet. read it first before writing to it.<<path>>",
"priority": "high",
"recommendation": "Investigate recurring error on <tool_use_error>file has not been read yet. read it first before writing to it.<<path>>",
"repo": "issue-facade",
"score": 290.0,
"signal_type": "recurring_error",
"title": "problem: recurring error"
},
{
"cross_flavor": false,
"pattern_key": "problem:recurring_error:<tool_use_error>file has not been read yet. read it first before writing to it.<<path>>",
"priority": "high",
"recommendation": "Investigate recurring error on <tool_use_error>file has not been read yet. read it first before writing to it.<<path>>",
"repo": "railiance-apps",
"score": 290.0,
"signal_type": "recurring_error",
"title": "problem: recurring error"
},
{
"cross_flavor": false,
"pattern_key": "problem:recurring_error:<tool_use_error>file has not been read yet. read it first before writing to it.<<path>>",
"priority": "high",
"recommendation": "Investigate recurring error on <tool_use_error>file has not been read yet. read it first before writing to it.<<path>>",
"repo": "state-hub",
"score": 290.0,
"signal_type": "recurring_error",
"title": "problem: recurring error"
},
{
"cross_flavor": false,
"pattern_key": "problem:recurring_error:<tool_use_error>file has not been read yet. read it first before writing to it.<<path>>",
"priority": "high",
"recommendation": "Investigate recurring error on <tool_use_error>file has not been read yet. read it first before writing to it.<<path>>",
"repo": "the-custodian",
"score": 290.0,
"signal_type": "recurring_error",
"title": "problem: recurring error"
},
{
"cross_flavor": false,
"pattern_key": "problem:recurring_error:<tool_use_error>file has not been read yet. read it first before writing to it.<<path>>",
"priority": "high",
"recommendation": "Investigate recurring error on <tool_use_error>file has not been read yet. read it first before writing to it.<<path>>",
"repo": "vergabe-teilnahme",
"score": 290.0,
"signal_type": "recurring_error",
"title": "problem: recurring error"
},
{
"cross_flavor": false,
"pattern_key": "problem:recurring_error:<tool_use_error>file has been modified since read, either by the user or by a linter. read it again before attempting to write it.<<path>>",
"priority": "medium",
"recommendation": "Investigate recurring error on <tool_use_error>file has been modified since read, either by the user or by a linter. read it again before attempting to write it.<<path>>",
"repo": "artifact-store",
"score": 78.0,
"signal_type": "recurring_error",
"title": "problem: recurring error"
},
{
"cross_flavor": false,
"pattern_key": "problem:recurring_error:<tool_use_error>file has been modified since read, either by the user or by a linter. read it again before attempting to write it.<<path>>",
"priority": "medium",
"recommendation": "Investigate recurring error on <tool_use_error>file has been modified since read, either by the user or by a linter. read it again before attempting to write it.<<path>>",
"repo": "issue-facade",
"score": 78.0,
"signal_type": "recurring_error",
"title": "problem: recurring error"
},
{
"cross_flavor": false,
"pattern_key": "problem:recurring_error:<tool_use_error>file has been modified since read, either by the user or by a linter. read it again before attempting to write it.<<path>>",
"priority": "medium",
"recommendation": "Investigate recurring error on <tool_use_error>file has been modified since read, either by the user or by a linter. read it again before attempting to write it.<<path>>",
"repo": "railiance-apps",
"score": 78.0,
"signal_type": "recurring_error",
"title": "problem: recurring error"
},
{
"cross_flavor": false,
"pattern_key": "problem:recurring_error:<tool_use_error>file has been modified since read, either by the user or by a linter. read it again before attempting to write it.<<path>>",
"priority": "medium",
"recommendation": "Investigate recurring error on <tool_use_error>file has been modified since read, either by the user or by a linter. read it again before attempting to write it.<<path>>",
"repo": "state-hub",
"score": 78.0,
"signal_type": "recurring_error",
"title": "problem: recurring error"
},
{
"cross_flavor": false,
"pattern_key": "problem:budget_overrun:tokens",
"priority": "medium",
"recommendation": "Read narrowly \u2014 target the region you need, not whole large files",
"repo": "artifact-store",
"score": 50.55,
"signal_type": "budget_overrun",
"title": "problem: budget overrun"
},
{
"cross_flavor": false,
"pattern_key": "problem:recurring_error:{",
"priority": "medium",
"recommendation": "Investigate recurring error on {",
"repo": "vergabe-teilnahme",
"score": 12.0,
"signal_type": "recurring_error",
"title": "problem: recurring error"
},
{
"cross_flavor": false,
"pattern_key": "problem:recurring_error:found <n> errors (<n> fixed, <n> remaining).",
"priority": "medium",
"recommendation": "Investigate recurring error on found <n> errors (<n> fixed, <n> remaining).",
"repo": "ops-bridge",
"score": 10.0,
"signal_type": "recurring_error",
"title": "problem: recurring error"
},
{
"cross_flavor": false,
"pattern_key": "problem:recurring_error:(note: edit also tried swapping \\uxxxx escapes and their characters; neither form matched, so the mismatch is likely elsewhere in old_string. re-read the file a",
"priority": "medium",
"recommendation": "Investigate recurring error on (note: edit also tried swapping \\uxxxx escapes and their characters; neither form matched, so the mismatch is likely elsewhere in old_string. re-read the file a",
"repo": "net-kingdom",
"score": 6.0,
"signal_type": "recurring_error",
"title": "problem: recurring error"
},
{
"cross_flavor": false,
"pattern_key": "problem:recurring_error:found <n> error (<n> fixed, <n> remaining).",
"priority": "medium",
"recommendation": "Investigate recurring error on found <n> error (<n> fixed, <n> remaining).",
"repo": "ops-bridge",
"score": 6.0,
"signal_type": "recurring_error",
"title": "problem: recurring error"
},
{
"cross_flavor": false,
"pattern_key": "problem:recurring_error:<n> failed, <n> passed in <n>.00s",
"priority": "medium",
"recommendation": "Investigate recurring error on <n> failed, <n> passed in <n>.00s",
"repo": "agentic-resources",
"score": 4.0,
"signal_type": "recurring_error",
"title": "problem: recurring error"
}
],
"window": {
"days": 30,
"since": "2026-05-08T17:14:00Z",
"until": "2026-06-07T17:14:00Z"
}
}

39
session_memory/retro/last_retro.md Normal file
View File

@@ -0,0 +1,39 @@
# Weekly Coding Retro (2026-05-08 → 2026-06-07)
_23 real sessions · generated 2026-06-07T17:14:00Z_
## Top improvement suggestions (cross-flavor first, ≤3 per repo)
- **net-kingdom** (high, score=54.0) [CROSS-FLAVOR]: cross-flavor problem: recurring error — Investigate recurring error on make: *** [makefile:<n>: fix-consistency] error <n>
- **activity-core** (high, score=13128.0): problem: tool thrash — Batch related shell work into one script, not many small Bash calls
- **artifact-store** (high, score=13128.0): problem: tool thrash — Batch related shell work into one script, not many small Bash calls
- **citation-evidence** (high, score=13128.0): problem: tool thrash — Batch related shell work into one script, not many small Bash calls
- **infospace-bench** (high, score=13128.0): problem: tool thrash — Batch related shell work into one script, not many small Bash calls
- **railiance-apps** (high, score=13128.0): problem: tool thrash — Batch related shell work into one script, not many small Bash calls
- **state-hub** (high, score=13128.0): problem: tool thrash — Batch related shell work into one script, not many small Bash calls
- **activity-core** (high, score=441.0): problem: schema thrash — Load the tool schemas you'll need once, up front
- **citation-evidence** (high, score=441.0): problem: schema thrash — Load the tool schemas you'll need once, up front
- **flex-auth** (high, score=441.0): problem: schema thrash — Load the tool schemas you'll need once, up front
- **infospace-bench** (high, score=441.0): problem: schema thrash — Load the tool schemas you'll need once, up front
- **ops-bridge** (high, score=441.0): problem: schema thrash — Load the tool schemas you'll need once, up front
- **activity-core** (high, score=290.0): problem: recurring error — Investigate recurring error on <tool_use_error>file has not been read yet. read it first before writing to it.<<path>>
- **citation-evidence** (high, score=290.0): problem: recurring error — Investigate recurring error on <tool_use_error>file has not been read yet. read it first before writing to it.<<path>>
- **infospace-bench** (high, score=290.0): problem: recurring error — Investigate recurring error on <tool_use_error>file has not been read yet. read it first before writing to it.<<path>>
- **issue-facade** (high, score=290.0): problem: recurring error — Investigate recurring error on <tool_use_error>file has not been read yet. read it first before writing to it.<<path>>
- **railiance-apps** (high, score=290.0): problem: recurring error — Investigate recurring error on <tool_use_error>file has not been read yet. read it first before writing to it.<<path>>
- **state-hub** (high, score=290.0): problem: recurring error — Investigate recurring error on <tool_use_error>file has not been read yet. read it first before writing to it.<<path>>
- **the-custodian** (high, score=290.0): problem: recurring error — Investigate recurring error on <tool_use_error>file has not been read yet. read it first before writing to it.<<path>>
- **vergabe-teilnahme** (high, score=290.0): problem: recurring error — Investigate recurring error on <tool_use_error>file has not been read yet. read it first before writing to it.<<path>>
- **artifact-store** (medium, score=78.0): problem: recurring error — Investigate recurring error on <tool_use_error>file has been modified since read, either by the user or by a linter. read it again before attempting to write it.<<path>>
- **issue-facade** (medium, score=78.0): problem: recurring error — Investigate recurring error on <tool_use_error>file has been modified since read, either by the user or by a linter. read it again before attempting to write it.<<path>>
- **railiance-apps** (medium, score=78.0): problem: recurring error — Investigate recurring error on <tool_use_error>file has been modified since read, either by the user or by a linter. read it again before attempting to write it.<<path>>
- **state-hub** (medium, score=78.0): problem: recurring error — Investigate recurring error on <tool_use_error>file has been modified since read, either by the user or by a linter. read it again before attempting to write it.<<path>>
- **artifact-store** (medium, score=50.55): problem: budget overrun — Read narrowly — target the region you need, not whole large files
- **vergabe-teilnahme** (medium, score=12.0): problem: recurring error — Investigate recurring error on {
- **ops-bridge** (medium, score=10.0): problem: recurring error — Investigate recurring error on found <n> errors (<n> fixed, <n> remaining).
- **net-kingdom** (medium, score=6.0): problem: recurring error — Investigate recurring error on (note: edit also tried swapping \uxxxx escapes and their characters; neither form matched, so the mismatch is likely elsewhere in old_string. re-read the file a
- **ops-bridge** (medium, score=6.0): problem: recurring error — Investigate recurring error on found <n> error (<n> fixed, <n> remaining).
- **agentic-resources** (medium, score=4.0): problem: recurring error — Investigate recurring error on <n> failed, <n> passed in <n>.00s
## Fleet snapshot
- infra-overhead median: 0.167
- error rate: 0.957 · schema-thrash: 7
- success rate: 1.0 · tokens p50: 250725

78
session_memory/retro/publish.py Normal file
View File

@@ -0,0 +1,78 @@
"""Publish the weekly retro (AGENTIC-WP-0010 T02).
The retro is published to the State Hub as a **read model** — a progress event of
``event_type=coding_retro`` whose ``detail`` carries the structured report. This is
exactly how ``daily-triage-report`` surfaces, and it is what activity-core's
``coding_retro`` resolver (ACTIVITY-WP-0008) reads. A local JSON + markdown report
is always written; the hub publish is best-effort and **degrades gracefully** when
the hub is unreachable.
"""
from __future__ import annotations
import json
import os
import urllib.request
from typing import Callable, Optional
DEFAULT_HUB = "http://127.0.0.1:8000"
def render_markdown(report: dict) -> str:
w = report.get("window", {})
lines = [
f"# Weekly Coding Retro ({w.get('since', '')[:10]} → {w.get('until', '')[:10]})",
f"_{report.get('n_sessions', 0)} real sessions · generated {report.get('generated_at', '')}_",
"",
"## Top improvement suggestions (cross-flavor first, ≤3 per repo)",
]
if not report.get("suggestions"):
lines.append("- (no recurring problems above threshold this week)")
for s in report.get("suggestions", []):
flag = " [CROSS-FLAVOR]" if s.get("cross_flavor") else ""
lines.append(f"- **{s['repo']}** ({s['priority']}, score={s['score']}){flag}: "
f"{s['title']} — {s['recommendation']}")
m = report.get("measure", {})
lines += ["", "## Fleet snapshot",
f"- infra-overhead median: {m.get('infra_overhead_share_median')}",
f"- error rate: {m.get('error_rate')} · schema-thrash: {m.get('schema_thrash_sessions')}",
f"- success rate: {m.get('success_rate')} · tokens p50: {m.get('tokens_p50')}"]
return "\n".join(lines)
def write_local(report: dict, json_path: str, md_path: Optional[str] = None) -> None:
os.makedirs(os.path.dirname(json_path) or ".", exist_ok=True)
with open(json_path, "w", encoding="utf-8") as fh:
json.dump(report, fh, indent=2, sort_keys=True)
fh.write("\n")
if md_path:
with open(md_path, "w", encoding="utf-8") as fh:
fh.write(render_markdown(report))
fh.write("\n")
def _http_post(url: str, payload: dict) -> None:
req = urllib.request.Request(url, data=json.dumps(payload).encode(),
headers={"Content-Type": "application/json"}, method="POST")
with urllib.request.urlopen(req, timeout=10) as r:
r.read()
def publish_to_hub(report: dict, *, base_url: str = DEFAULT_HUB,
poster: Optional[Callable[[str, dict], None]] = None) -> bool:
"""POST the retro as an event_type=coding_retro progress event. Best-effort."""
poster = poster or _http_post
n = report.get("n_sessions", 0)
k = len(report.get("suggestions", []))
payload = {
"event_type": "coding_retro",
"author": "helix-forge",
"summary": f"Weekly coding retro: {k} ranked suggestions across "
f"{report.get('window', {}).get('days', 7)} days ({n} sessions).",
"detail": report,
}
try:
poster(f"{base_url.rstrip('/')}/progress/", payload)
return True
except Exception:
return False

86
tests/test_retro_build.py Normal file
View File

@@ -0,0 +1,86 @@
"""Weekly retro report tests (AGENTIC-WP-0010 T01)."""
import os
import sys
sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
from session_memory.curate.catalog import Catalog # noqa: E402
from session_memory.curate.schema import Resolution, SolutionPattern # noqa: E402
from session_memory.retro.build import weekly_retro # noqa: E402
def _digest(uid, repo, ts, flavor="claude", retries=5):
return {
"session_uid": uid, "flavor": flavor, "repo": repo, "outcome": "fail",
"started_at": ts, "event_count": 40,
"first_prompt": "Fix the failing build and retry the suite",
"cost": {"input_tokens": 100, "output_tokens": 10},
"tool_histogram": {"Bash": 20, "Edit": 12, "Read": 8},
"markers": {"errors": 0, "retries": retries, "test_runs": 0},
"error_snippets": [],
}
def test_window_excludes_old_sessions():
digs = [
_digest("claude:a", "r1", "2026-06-01T10:00:00Z"),
_digest("claude:b", "r1", "2026-06-02T10:00:00Z"),
_digest("claude:old", "r1", "2026-01-01T10:00:00Z"), # outside window
]
r = weekly_retro(digs, since="2026-05-30T00:00:00Z", until="2026-06-08T00:00:00Z")
assert r["n_sessions"] == 2
assert r["window"]["days"] == 7
def test_retry_storm_becomes_suggestion():
digs = [_digest(f"claude:{i}", "r1", "2026-06-0{}T10:00:00Z".format(i + 1))
for i in range(2)]
r = weekly_retro(digs, since="2026-05-30T00:00:00Z", until="2026-06-08T00:00:00Z")
s = r["suggestions"]
assert s and s[0]["repo"] == "r1"
assert s[0]["signal_type"] == "retry_storm"
assert "Investigate" in s[0]["recommendation"] # no catalog -> default
def test_recommendation_from_catalog(tmp_path):
cat = Catalog(str(tmp_path / "catalog"))
key = "problem:retry_storm:retries"
cat.upsert(SolutionPattern(
id=SolutionPattern.make_id(key), name="Retry storm", version="1.0.0",
polarity="problem", problem="repeated retries",
resolutions=[Resolution(summary="Stop and diagnose before retrying")]))
digs = [_digest(f"claude:{i}", "r1", "2026-06-0{}T10:00:00Z".format(i + 1)) for i in range(2)]
r = weekly_retro(digs, catalog=cat, since="2026-05-30T00:00:00Z", until="2026-06-08T00:00:00Z")
assert r["suggestions"][0]["recommendation"] == "Stop and diagnose before retrying"
def test_caps_three_per_repo():
# five distinct problem signals in one repo -> capped at 3
digs = []
for i in range(2):
d = _digest(f"claude:{i}", "r1", "2026-06-0{}T10:00:00Z".format(i + 1))
d["markers"] = {"errors": 5, "retries": 5, "test_runs": 0, "human_interventions": 0}
d["tool_histogram"] = {"Bash": 120, "ToolSearch": 9,
"mcp__state-hub__x": 30, "Edit": 5}
d["outcome"] = "abandoned"
digs.append(d)
r = weekly_retro(digs, since="2026-05-30T00:00:00Z", until="2026-06-08T00:00:00Z")
per_repo = [s for s in r["suggestions"] if s["repo"] == "r1"]
assert len(per_repo) <= 3
def test_cross_flavor_ranks_first():
digs = [
_digest("claude:a", "r1", "2026-06-01T10:00:00Z", flavor="claude"),
_digest("grok:b", "r2", "2026-06-02T10:00:00Z", flavor="grok"),
]
r = weekly_retro(digs, since="2026-05-30T00:00:00Z", until="2026-06-08T00:00:00Z")
assert r["suggestions"][0]["cross_flavor"] is True
assert r["suggestions"][0]["priority"] == "high"
def test_includes_measure_snapshot():
digs = [_digest(f"claude:{i}", "r1", "2026-06-0{}T10:00:00Z".format(i + 1)) for i in range(2)]
r = weekly_retro(digs, since="2026-05-30T00:00:00Z", until="2026-06-08T00:00:00Z")
assert r["measure"]["n_sessions"] == 2

63
tests/test_retro_entrypoint.py Normal file
View File

@@ -0,0 +1,63 @@
"""Retro entrypoint tests (AGENTIC-WP-0010 T03)."""
import json
import os
import sys
sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
from session_memory.core.store import Store # noqa: E402
from session_memory.retro.__main__ import main, run_retro # noqa: E402
def _digest(uid, repo, ts, retries=5):
return {
"session_uid": uid, "flavor": "claude", "repo": repo, "outcome": "fail",
"started_at": ts, "event_count": 40,
"first_prompt": "Fix the failing build and retry the suite repeatedly",
"cost": {"input_tokens": 100, "output_tokens": 10},
"tool_histogram": {"Bash": 20, "Edit": 12, "Read": 8},
"markers": {"errors": 0, "retries": retries, "test_runs": 0},
"error_snippets": [],
}
def _config(tmp_path):
store = tmp_path / ".store"
toml = tmp_path / "config.toml"
toml.write_text(
f'[store]\ndb_path="{store / "m.db"}"\nblob_dir="{store / "blobs"}"\ncursor="{store / "c.json"}"\n'
f'[curate]\ncatalog_dir="{tmp_path / "catalog"}"\n'
f'[retro]\nwindow_days=7\nreport_json="{tmp_path / "r.json"}"\nreport_md="{tmp_path / "r.md"}"\n')
st = Store(str(store / "m.db"), str(store / "blobs"))
st.write_digest("claude:a", _digest("claude:a", "r1", "2026-06-01T10:00:00Z"))
st.write_digest("claude:b", _digest("claude:b", "r1", "2026-06-02T10:00:00Z"))
st.close()
return str(toml), tmp_path
def test_run_retro_over_store(tmp_path):
from session_memory.ingest import load_config
cfg_path, _ = _config(tmp_path)
rep = run_retro(load_config(cfg_path), since="2026-05-30T00:00:00Z", until="2026-06-08T00:00:00Z")
assert rep["n_sessions"] == 2
assert rep["suggestions"]
def test_main_writes_report_files(tmp_path, capsys):
cfg_path, tp = _config(tmp_path)
rc = main(["--config", cfg_path, "--since", "2026-05-30T00:00:00Z",
"--until", "2026-06-08T00:00:00Z"])
assert rc == 0
assert os.path.exists(str(tp / "r.json")) and os.path.exists(str(tp / "r.md"))
assert "Weekly Coding Retro" in capsys.readouterr().out
def test_main_json(tmp_path, capsys):
cfg_path, _ = _config(tmp_path)
rc = main(["--config", cfg_path, "--since", "2026-05-30T00:00:00Z",
"--until", "2026-06-08T00:00:00Z", "--json"])
assert rc == 0
data = json.loads(capsys.readouterr().out)
assert data["report"]["n_sessions"] == 2
assert data["published"] is None # no --publish

62
tests/test_retro_publish.py Normal file
View File

@@ -0,0 +1,62 @@
"""Retro publish tests (AGENTIC-WP-0010 T02)."""
import json
import os
import sys
sys.path.insert(0, os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
from session_memory.retro.publish import ( # noqa: E402
publish_to_hub,
render_markdown,
write_local,
)
def _report():
return {
"window": {"since": "2026-06-01T00:00:00Z", "until": "2026-06-08T00:00:00Z", "days": 7},
"generated_at": "2026-06-08T19:00:00Z", "n_sessions": 12,
"suggestions": [
{"repo": "state-hub", "title": "schema thrash", "recommendation": "front-load schemas",
"priority": "high", "score": 632.0, "cross_flavor": False, "signal_type": "schema_thrash"},
],
"measure": {"infra_overhead_share_median": 0.117, "error_rate": 0.96,
"schema_thrash_sessions": 8, "success_rate": 1.0, "tokens_p50": 250725},
}
def test_render_markdown():
md = render_markdown(_report())
assert "Weekly Coding Retro" in md
assert "**state-hub**" in md and "front-load schemas" in md
assert "infra-overhead median: 0.117" in md
def test_write_local_json_and_md(tmp_path):
jp = str(tmp_path / "out" / "retro.json")
mp = str(tmp_path / "out" / "retro.md")
write_local(_report(), jp, mp)
assert json.load(open(jp))["n_sessions"] == 12
assert "Weekly Coding Retro" in open(mp).read()
def test_publish_calls_poster_with_coding_retro_event():
captured = {}
def poster(url, payload):
captured["url"] = url
captured["payload"] = payload
ok = publish_to_hub(_report(), base_url="http://hub", poster=poster)
assert ok is True
assert captured["url"] == "http://hub/progress/"
assert captured["payload"]["event_type"] == "coding_retro"
assert captured["payload"]["detail"]["n_sessions"] == 12
def test_publish_degrades_gracefully_on_failure():
def boom(url, payload):
raise OSError("hub down")
assert publish_to_hub(_report(), poster=boom) is False

76
workplans/AGENTIC-WP-0010-weekly-retro.md Normal file
View File

@@ -0,0 +1,76 @@
---
id: AGENTIC-WP-0010
type: workplan
title: "Coding Session Memory — Weekly Retro entrypoint + hub publish"
domain: helix_forge
repo: agentic-resources
status: finished
owner: codex
topic_slug: helix-forge
created: "2026-06-07"
updated: "2026-06-07"
state_hub_workstream_id: "6b9816e4-65bc-4fc7-b8e1-33f4edd51e7a"
---
# Coding Session Memory — Weekly Retro entrypoint + hub publish
The **analysis half** of a weekly coding retrospection. A windowed retro runs
detect + measure over the previous week, ranks the **top-3 improvement
suggestions per repo** (impact × frequency, cross-flavor first; recommendations
pulled from the Pattern Catalog), and **publishes the ranked result to the State
Hub as a read model** (an `event_type=coding_retro` progress event, mirroring how
`daily-triage-report` publishes).
This is the dependency that activity-core's weekly schedule consumes
(`activity-wp-0008` — *Weekly Coding Retrospection schedule*). Keeping the analysis
here and publishing to the hub keeps activity-core decoupled from the
workstation-local session store.
## Windowed Weekly Retro Report (top-3 per repo)
```task
id: AGENTIC-WP-0010-T01
status: done
priority: high
state_hub_task_id: "34d30250-c0d3-4837-81c7-1c858c2ee801"
```
`retro/build.py`: window digests by date (last N days), run
`extract_signals` + `cluster` over the window, explode problem patterns per repo,
rank by score and cap at **3 per repo**. Attach a recommendation per suggestion
from the Pattern Catalog (lookup by pattern key → first resolution) with a sensible
default. Include a fleet measure snapshot for context. Pure function over digests;
unit-tested.
## Publish Retro to the Hub + Local Report
```task
id: AGENTIC-WP-0010-T02
status: done
priority: high
state_hub_task_id: "cbe1288a-ce51-48c0-b741-adf4a6cbce3a"
```
Publish the ranked retro to the State Hub as a read model: POST a progress event
(`event_type=coding_retro`) with the structured report (`suggestions[]`, window,
`generated_at`) in `detail`. Also write a local JSON + markdown report. **Graceful
degrade** when the hub is unreachable (write local, skip publish). Hub URL under
`[retro]` in `config.toml`.
## Retro Entrypoint + Tests + Live Verify
```task
id: AGENTIC-WP-0010-T03
status: done
priority: medium
state_hub_task_id: "af540220-58dd-4cf5-a9dc-6db4b995fa08"
```
`python -m session_memory.retro [--window-days 7] [--publish] [--json]`: windowed
retro → ranked top-3 per repo → optional hub publish + local report. Document in
`session_memory/README.md`. Live verify over the real local sessions. After
workplan updates, notify the operator to run from `~/state-hub`:
```bash
make fix-consistency REPO=agentic-resources
```