feat(lessons): the REST door a human can actually reach (#3734)
CI & Build / Python lint (push) Successful in 2s
CI & Build / Plugin hooks (push) Successful in 8s
CI & Build / integration (push) Successful in 48s
CI & Build / TypeScript typecheck (push) Successful in 52s
CI & Build / Python tests (push) Successful in 1m33s
CI & Build / Build & push image (push) Successful in 23s

Step 7, part two. Milestone 385 built the lesson kind through the MCP tools,
which is the agent's surface. The Vue app speaks REST, so a lesson was a
record a person could not create, read, edit or retire — rule 27 failing at
the door rather than in the view.

`/api/lessons` now offers list, create, read, update and trash, plus
`/api/lessons/taught-by/<id>` — the reverse of `learned_from`, which the task
body calls the direction that gets forgotten and arguably the more useful one:
a reader opening an old issue wants to know what was learned from it, and
until now the relation was only navigable from the lesson's side.

`lessons_taught_by` reads `data[taught_by]` through `path_exists`, the same
jsonpath dialect the snippet location lookup uses, so both reverse lookups hit
the GIN index (0070) the same way rather than scanning bodies. Share-aware via
`readable_notes_clause`: it renders beside a record the caller can already
see, so a lesson shared with them belongs there exactly as their own does.

THE TRIGGER IS REFUSED WHEN EMPTY, at create and at update. This is the one
place the door is not a thin wrapper, and it is deliberate: the service will
store a triggerless lesson quite happily — it saves, reads correctly in every
listing, and never surfaces. There is nothing to notice afterwards, because it
looks exactly like a lesson that works. Better to refuse it than to hand back
a record that looks finished. The refusal says why, so the next reader does
not take it for a nag and delete it.

`lesson_to_dict` moves into the service and the MCP tool's `_to_dict` becomes
an alias for it. Both doors now return one shape — a payload spelled once per
door answers the two of them differently the first time a field is added — and
both compose through `services/lessons.py`, so a lesson written from the web
ranks identically to one written by an agent. The document IS what ranks, so
that parity is the whole reason the door is thin.

The dedup gate matches the MCP path: two lessons under one trigger compete in
a single ranked list for one reserved slot, so a duplicate here displaces
rather than merely clutters.

NOT DONE YET: this is the door, not the UI. #3734 stays in_progress until the
Vue views, the router entries, the Knowledge browse badge and the both-ways
sources panel exist — rule 27 is about the operator being able to touch it.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01821k5B3Ysecp9fNYs92Kuy
This commit is contained in:
2026-09-19 14:02:57 -04:00
co-authored by Claude Opus 5
parent 1252d0e305
commit d36d68a20f
5 changed files with 578 additions and 15 deletions
+251
View File
@@ -0,0 +1,251 @@
"""REST routes for lessons — a transferable insight, retrievable by situation.
A lesson is a note with note_type='lesson' (see services/lessons.py). These
routes feed the web UI; the MCP tools (mcp/tools/lessons.py) are the
agent-facing surface. Both go through services/lessons.py, so the compose/parse
contract and the `data` mirror live in one place — the same division snippets
use, and for the same reason: two doors that each compose a lesson would
compose it two ways, and the document IS what ranks.
ACL (rule #78): reads and writes of a single lesson resolve through the
share-aware `get_lesson` / `can_write_note`, and writes are performed as the
OWNER so a shared editor isn't rejected by the owner-scoped service —
mirroring routes/snippets.py and routes/notes.py.
WHY THE TRIGGER IS A NAMED FIELD HERE TOO. The web editor could have posted a
body and let the service parse it. It doesn't, because the evidence behind
this kind (step 1) is that a trigger gets filled when a door ASKS for it by
name — the snippet corpus is at 100% on its trigger with no guard anywhere,
because a service composes the title from a parameter. A form that offered one
markdown box would be the option milestone 385 rejected, wearing a different
hat.
"""
import logging
from quart import Blueprint, jsonify, request
from scribe.auth import get_current_user_id, login_required
from scribe.routes.utils import not_found, parse_pagination
from scribe.services import dedup as dedup_svc
from scribe.services import knowledge as knowledge_svc
from scribe.services import lessons as lessons_svc
from scribe.services import systems as systems_svc
from scribe.services import trash as trash_svc
from scribe.services.access import (
can_write_note,
describe_provenance,
label_shared_items,
)
from scribe.services.note_usage import empty_usage, record_pulled, usage_for_notes
logger = logging.getLogger(__name__)
lessons_bp = Blueprint("lessons", __name__, url_prefix="/api/lessons")
@lessons_bp.route("", methods=["GET"])
@login_required
async def list_lessons_route():
"""The kind enumerated, rather than only what a query resembles.
Semantic search is how a lesson REACHES a session; this is how a person
sees what exists at all. Each row carries its trigger, because a list of
lessons without them is a list of claims with the half that says when each
one matters left off.
"""
uid = get_current_user_id()
q = request.args.get("q") or None
tag = request.args.get("tag", "")
try:
project_id = int(request.args.get("project_id", 0) or 0) or None
except (TypeError, ValueError):
project_id = None
limit, offset = parse_pagination(default_limit=24, max_limit=100)
items, total = await knowledge_svc.query_knowledge(
user_id=uid,
note_type=lessons_svc.LESSON_NOTE_TYPE,
tags=[tag] if tag else [],
sort="modified",
q=q,
limit=limit,
offset=offset,
project_id=project_id,
)
return jsonify({
"lessons": await label_shared_items(uid, items),
"total": total,
})
@lessons_bp.route("/taught-by/<int:record_id>", methods=["GET"])
@login_required
async def lessons_taught_by_route(record_id: int):
"""The lessons drawn FROM one record — the reverse of `learned_from`.
Registered ABOVE the `/<int:lesson_id>` routes on purpose: Quart matches
in registration order, and `taught-by` would otherwise never be reached
if the converter ever widened. The same ordering snippets' `/duplicates`
route documents.
This is the direction that gets forgotten. A reader opening an old issue
wants to know what was learned from it, and without this the relation is
only navigable from the lesson's side.
"""
uid = get_current_user_id()
notes = await lessons_svc.lessons_taught_by(uid, record_id)
return jsonify({
"lessons": [lessons_svc.lesson_to_dict(n) for n in notes],
"taught_by": record_id,
})
@lessons_bp.route("", methods=["POST"])
@login_required
async def create_lesson_route():
uid = get_current_user_id()
data = await request.get_json() or {}
what = (data.get("what") or "").strip()
when_to_apply = (data.get("when_to_apply") or "").strip()
if not what:
return jsonify({"error": "what is required"}), 400
# The trigger is not optional at this door even though the service will
# store a lesson without one. A lesson with no trigger saves, reads
# correctly in every listing, and never surfaces — there is nothing to
# notice afterwards, which is exactly why the form has to refuse it here
# rather than leave the writer a record that looks finished.
if not when_to_apply:
return jsonify({
"error": "when_to_apply is required",
"detail": (
"A lesson is found by the SITUATION it applies to. Without a "
"trigger it still saves and still reads correctly, and it "
"never reaches anyone — so it is refused here rather than "
"stored as a record that looks finished."
),
}), 400
project_id = data.get("project_id") or None
learned_from = data.get("learned_from") or []
# The same near-duplicate gate the MCP create path applies. Two lessons
# under one trigger compete in a single ranked list for one reserved slot,
# so the duplicate does not merely clutter — it displaces.
if not data.get("force"):
title, body = lessons_svc.lesson_document(
what, when_to_apply, data.get("insight", ""), learned_from,
)
dup = await dedup_svc.find_duplicate_note(
uid, title, body,
project_id=project_id,
is_task=False,
note_type=lessons_svc.LESSON_NOTE_TYPE,
)
if dup is not None:
return jsonify(dedup_svc.duplicate_response(dup, "lesson")), 409
note = await lessons_svc.create_lesson(
uid,
what=what,
when_to_apply=when_to_apply,
insight=data.get("insight", ""),
learned_from=learned_from,
tags=data.get("tags"),
project_id=project_id,
)
if data.get("system_ids") is not None:
await systems_svc.set_record_systems(uid, note.id, data["system_ids"])
out = lessons_svc.lesson_to_dict(note)
out["systems"] = [
s.to_dict() for s in await systems_svc.list_record_systems(uid, note.id)
]
return jsonify(out), 201
@lessons_bp.route("/<int:lesson_id>", methods=["GET"])
@login_required
async def get_lesson_route(lesson_id: int):
uid = get_current_user_id()
note = await lessons_svc.get_lesson(uid, lesson_id)
if note is None:
return not_found("Lesson")
out = lessons_svc.lesson_to_dict(note)
# As the OWNER: a shared reader isn't scoped to the owner's project, so
# their own id would come back empty (the write-as-owner pattern this
# module already uses, read side).
out["systems"] = [
s.to_dict()
for s in await systems_svc.list_record_systems(note.user_id, lesson_id)
]
out.update(await describe_provenance(uid, note))
out["usage"] = (await usage_for_notes([lesson_id])).get(
lesson_id, empty_usage()
)
# Opening the detail view IS a pull — the operator chose to look. Tagged
# apart from the MCP sources so "an agent was handed it" and "a human read
# it" stay distinguishable; they mean different things for pruning (#2085).
record_pulled(user_id=uid, note_id=lesson_id, source="rest_lesson")
return jsonify(out)
@lessons_bp.route("/<int:lesson_id>", methods=["PATCH"])
@login_required
async def update_lesson_route(lesson_id: int):
uid = get_current_user_id()
note = await lessons_svc.get_lesson(uid, lesson_id)
if note is None:
return not_found("Lesson")
if not await can_write_note(uid, lesson_id):
return jsonify({"error": "Permission denied"}), 403
owner_uid = note.user_id
data = await request.get_json() or {}
# Partial update: only keys present in the payload change, and the service
# re-composes title, body and mirror from the merged set — so a form that
# sends one field cannot leave the halves of the document disagreeing.
kwargs = {
k: data[k]
for k in ("what", "when_to_apply", "insight", "learned_from", "tags")
if k in data
}
# An empty trigger would save and silently stop the lesson surfacing, so
# clearing it is refused for the same reason creating without one is.
if "when_to_apply" in kwargs and not (kwargs["when_to_apply"] or "").strip():
return jsonify({
"error": "when_to_apply cannot be cleared",
"detail": (
"A lesson with no trigger never surfaces, and nothing about "
"the stored record would show it. Rewrite the trigger rather "
"than emptying it."
),
}), 400
updated = await lessons_svc.update_lesson(owner_uid, lesson_id, **kwargs)
if updated is None:
return not_found("Lesson")
if data.get("system_ids") is not None:
await systems_svc.set_record_systems(
owner_uid, lesson_id, data["system_ids"]
)
out = lessons_svc.lesson_to_dict(updated)
out["systems"] = [
s.to_dict()
for s in await systems_svc.list_record_systems(owner_uid, lesson_id)
]
return jsonify(out)
@lessons_bp.route("/<int:lesson_id>", methods=["DELETE"])
@login_required
async def delete_lesson_route(lesson_id: int):
"""Trash, not erase — recoverable from the trash like every other kind."""
uid = get_current_user_id()
note = await lessons_svc.get_lesson(uid, lesson_id)
if note is None:
return not_found("Lesson")
if not await can_write_note(uid, lesson_id):
return jsonify({"error": "Permission denied"}), 403
batch_id = await trash_svc.delete(note.user_id, "note", lesson_id)
if batch_id is None:
return not_found("Lesson")
return jsonify({"deleted": lesson_id, "deleted_batch_id": batch_id})