concinno-skills-session-search
Cross-session search for the Concinno agent. A long-running agent forgets prior sessions because each Claude Code transcript is isolated. This sub-package gives the agent a queryable SQLite FTS5 index over its own past sessions plus optional Haiku-generated summaries, so it can answer "did we already debug this?" / "what did the last KILL 10 wave decide?" / "where did we put the audit plan?" inside one tool-use round.
Install
pip install concinno-skills-session-search # core
pip install 'concinno-skills-session-search[anthropic]' # + Haiku summaries
The core install is stdlib-only — sqlite3 ships with CPython.
The anthropic SDK is optional; without it, the summariser falls back
to a truncate-first-N-chars heuristic so the import path never
hard-fails.
Quick start
from concinno_skills_session_search import SessionIndex
idx = SessionIndex() # ~/.concinno/session_search.sqlite by default
idx.index_session(
session_id="2026-04-28-W4-fast-followers",
transcript_text="we shipped HP4 cross-session search ...",
started_at=1714305600.0,
ended_at=1714312800.0,
)
for hit in idx.search("HP4 cross-session"):
print(hit.session_id, hit.summary[:120])
Lifecycle wiring
When Concinno main >=4.6.0 is installed, the three lifecycle hooks run automatically:
on_session_start— read-only no-op (the agent reads on demand).on_stop— index this session's transcript with a best-effort Haiku summary; ImportError / API errors are swallowed.on_session_end—idx.prune(days=90)enforces 90-day retention.
See concinno.hooks.on_session_start / on_stop for the wiring site.
CLI
concinno-session-search --query "audit plan" --limit 5
concinno-session-search --count
concinno-session-search --prune-days 90
concinno-session-search --reindex --session-id S1 --transcript-file t.txt
--format=json is supported on every subcommand.
Environment
| Variable | Purpose |
|---|---|
CONCINNO_SESSION_SEARCH_DB |
Override ~/.concinno/session_search.sqlite. |
ANTHROPIC_API_KEY |
Read by the optional Haiku summariser. |
License
AGPL-3.0-or-later. Copyright (C) 2026 Chen Syuan Wang (王晨宣 / AI King).
Metadata
Release files for concinno-skills-session-search 0.2.1
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| concinno_skills_session_search-0.2.1.tar.gz | 33.1 kB | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| concinno_skills_session_search-0.2.1-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 60.7 kB
Release files / concinno_skills_session_search-0.2.1.tar.gz
| Download URL | concinno_skills_session_search-0.2.1.tar.gz |
|---|---|
| Size | 33.1 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
96ee25b8d0da529d62d71d8487b1f0538e708a042dde75e95ba36bcf6af2a1d2
|
|
BLAKE2b-256 checksum How to use checksums |
68ff837b6503942a863b8ee648fb823255739ab8f35b70edb8f29e2674bb4b41
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.11.9
|
Release files / concinno_skills_session_search-0.2.1-py3-none-any.whl
| Download URL | concinno_skills_session_search-0.2.1-py3-none-any.whl |
|---|---|
| Size | 27.6 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
4d4b86afe47a3dec4372b824fd8d474f01af0e0871d25cd16efb78ee0a1ad818
|
|
BLAKE2b-256 checksum How to use checksums |
5cb3399b3e1ad1b1e7108c90d6c30f350b65df2f9427e6119ea98f26dba683f4
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.11.9
|