Grant's real-phone testing surfaced seven items; this lands six (the seventh,
in-app camera card intake, is planned in docs/handoffs/in-app-card-intake-plan.md).
CRM half — ships in the s9pk (v0.1.0:99):
- Intake fuzzy match no longer over-indexes on generic firm words. _name_similarity
now compares DISTINCTIVE tokens only (generic descriptors — "Investment Group",
"Capital", "Family Office" — stripped via _GENERIC_ORG_WORDS) for both the difflib
ratio and the Jaccard, so "Fortitude Investment Group" stops surfacing Aether/Russell
while "Aether Capital" still surfaces "Aether Investment Group". +2 regression cases.
- Mobile grid "Last contact"/staleness sort is reversible. SortSheet gains opt-in
dir/onToggleDir; other surfaces (Contacts/Pipeline) are untouched.
- Mobile "Edit investor" prefills a contact's saved email. GET /api/fundraising/state
heals a blank grid pill email from the linked classic contact
(fundraising_contacts.contact_id -> contacts.email), fill-only, by pill order then
name; the next one-row save persists it. +test_grid_email_heal.py.
- Mobile quick-log pencil icon renders. iOS collapses a sole, centered, attribute-only
-sized flex-child <svg>; .quicklog-btn svg now gets explicit CSS width/height + flex:none
(the pattern the working bottom-tab/sort-pill icons use). The v97 fix only changed color.
Matrix intake bot — ships on the Spark (bot-only, NOT the s9pk):
- Approve/reject now redacts the whole intake thread (card + ack + main-timeline nudge +
the user's own photo/note), mirroring the email-review room; redact_thread takes the
room as an arg and matches replies by m.thread OR m.in_reply_to (so the nudge clears).
No more in-Matrix confirmation after a commit (the thread vanishing is the ack).
Needs the bot to hold a redact/moderator power level in the intake room.
- New one-time backend/matrix_intake/redact_intake.py clears the room's pre-existing
backlog (dry-run default; --apply).
Tests 42/42 green; frontend render-smoke green. Frontend fixes are inspection + render
-smoke verified (on-device confirm pending); the bot redaction is live-smoke only.
Completes business-card contact capture. The transcription prompt now labels
Phone/Mobile/Fax on separate lines, and the extractor maps an office/main number ->
phone and a cell -> mobile, never a fax. Both carry the same digit-in-source
integrity rule as email/LinkedIn: a number is kept only if its digits literally
appear in the source (or, on revise, the instruction) -- never minted. The proposal
card shows Phone + Mobile and they're editable (aliases phone/tel/office, mobile/cell).
Server: _upsert_contact_from_fundraising now accepts contact.phone + contact.mobile
and writes them to the canonical contact record (contact-level, not grid pills),
shipped in s9pk v0.1.0:98. No schema change -- the contacts columns already exist.
41/41 backend suite green + the matrix_intake units; card flow end-to-end is live-smoke.
The card transcription prompt now reads emails/URLs/phones character-by-character,
explicitly forbids autocompleting toward a plausible domain (the mara.com ->
marac.com failure), and emits labeled lines (which also feeds the field extractor
cleaner input).
The extractor gains city + linkedin_url. city is a plain field (low-harm if wrong;
the human sees it on the card). linkedin_url follows the email-integrity rule: kept
only if it literally appears in the source / a revise instruction, never minted -- a
wrong profile URL points at the wrong person. Both flow to the contact via the
existing log-communication upsert (city also syncs to the grid contact pill).
Phone is intentionally NOT included yet: the bot's write path can't store it until a
small server-side change lands (next s9pk). See the matrix-intake guide.
The in-thread approval handler already routes any reply that isn't yes/no/
edit-grammar through local Qwen (parse.revise), but the card copy only mentioned
'edit field=value', so the natural-language path was undiscoverable. Lead with
plain-words edits; the deterministic field=value fast-path still works.
The intake bot now accepts a photo of a business card in the intake room and
turns it into the same new-investor proposal a typed note would. The only new
step is image -> text; everything downstream (parse, fuzzy match, in-thread
approval, log-communication write) is reused unchanged.
M3 was deferred only because Spark Control had no vision model. That blocker is
gone: the daily-driver Qwen is vision-capable under the same model id, and the
gateway forwards OpenAI multimodal content untouched, so no gateway/server/s9pk
change is needed -- this ships bot-only (git pull + rebuild on the Spark).
Transcribe-then-reuse (not vision-straight-to-JSON) is deliberate: the
transcription becomes the source text the email-integrity rule checks against,
so a mis-read address can't reach the CRM unapproved -- same guarantee as the
text path. Card commits tag source="matrix_card" for the audit log.
- llm.chat_vision: multimodal /v1/chat/completions, same model, same gateway
- spark.transcribe_card: faithful card->text, "" on a non-card (NONE sentinel)
- bot.on_image/handle_card: download image, transcribe, hand to handle_intake
- crm_client: source provenance overridable via the proposal's _source key
- tests: test_spark.py + a provenance case; 41/41 suite green
Read-only natural-language query over the curated nl_query endpoint, answered
in-thread. Two entry points (room-per-purpose model): a dedicated Q&A room
(MATRIX_QUERY_ROOM) where every top-level message is a question, plus the
?/@bot trigger in the intake room as a cross-room convenience. Both routes hit
the same handle_query -> crm_client.nl_query -> POST /api/query/nl; translation
runs on the box's local model, nothing leaves the box, and there is no write
path so no approval gate applies.
Pure logic (trigger parsing, answer rendering) in query.py with offline tests;
async room wiring in bot.py (live-smoke only, per the bot's convention).
Bot-side only, ships on the Spark via git pull + restart. Depends on the
box-side /api/query/nl endpoint, which lands with the v93 s9pk (reminders + W2):
until v93 is installed the Q&A surface 404s, so the bot deploy is staged to
follow that install.
The bot was granted a redact/mod power level in the review room, so it can now
clear a resolved thread entirely, not just the card: redact_thread redacts the
card root then scans recent history for its m.thread replies (the human's yes/no
+ any bot messages) and redacts those too, so decided threads drop out of the
threads view, not only the main timeline. Drops the in-thread confirmation on a
successful decision (the thread clearing is the ack; a confirmation would keep
the thread alive). redact_resolved.py extended to also clear replies of already-
resolved threads for the one-time backfill. Bot-only; no s9pk change.
Cards decided before the auto-redact behavior shipped are already 'closed' in
the CRM, so the bot's to_close sweep never redacts them. redact_resolved.py walks
the review room, keeps cards still pending (CRM 'open' list), and redacts the
rest. Dry-run by default; --apply to act. Run via docker compose on the Spark.
Three post-smoke refinements to the Matrix email-proposal review:
1. Dash separators (bot): every card/reply is framed with a dash rule top and
bottom so threads stop bleeding together vertically on mobile.
2. Remove decided threads (bot): on a conclusive approve/dismiss from either
surface, the bot redacts the card (client.room_redact) so the room clears
down to only undecided items. Redacting the bot's own card needs no power;
the web->Matrix path now redacts instead of posting a closure note.
3. Clearer note wording (server v91 + bot): the proposed grid note now names who
emailed whom -- "{teammate} emailed {investor}" (outbound) / "{sender} emailed
the team" (inbound) -- instead of an ambiguous "Sent"/"Received". Outbound
detection also matches our corporate domain (public providers excluded), so a
teammate's mail from a non-enrolled @ten31.xyz address no longer reads as
"Received". Going-forward only; no schema change. The card drops its bare
direction label since the note now carries the relationship.
Tests updated; 30/30 green, render-smoke green.
room_send to a room the bot is only invited to (not joined) fails M_FORBIDDEN;
join explicitly on startup (idempotent if already a member). Bot-only change —
ships via the Spark git pull, no s9pk bump.
The email-capture "proposed grid notes" gain two review surfaces:
1. Inline source email — each proposed-note card on the Email Capture page
gets a "View email" toggle that lazily fetches the existing
GET /api/email/detail and shows from/to/cc/date/subject + scrollable body,
so a reviewer can judge the note against the email it was drafted from.
2. CRM->Matrix review bridge — the CRM (box, stdlib, no matrix-nio) can't post
to Matrix, so the intake bot (Spark) PULLS: GET /api/intake/email-proposals
returns to_post/open/to_close work-lists; the bot posts a review card
(metadata + snippet + draft note) to a dedicated review room
(MATRIX_EMAIL_REVIEW_ROOM) and relays in-thread yes / no / NL-edit
(POST .../{id}/decide, note revised via local Qwen). Decisions sync both
ways: web decide -> bot announces + closes the thread; Matrix decide -> the
web panel's ~25s poll clears the card. State lives CRM-side in the new
email_proposal_matrix side row (email-integration migration 0003, additive
+ idempotent CREATE TABLE IF NOT EXISTS), so it survives a bot restart.
Adds a 'bot' role (authenticated, never admin; require_bot_or_admin) to gate
the email-proposal endpoints rather than handing the bot full admin — the
principled base for the coming agentic capabilities. Role controls reach;
the draft->approve gate still controls autonomy (a human approves every write).
Deploy split: endpoints + migration + role + frontend ship in the s9pk; the
bot poll loop + review-room handling ship on the Spark. The bot's CRM user
must be flipped member->bot and joined to the review room (one-time).
Tests: backend/test_email_proposal_matrix.py + matrix_intake/test_email_proposals.py
(30/30 suite green, render-smoke green, migration verified twice on a DB copy).
Local-smoke found "jonathan is chatting with wyoming" extracted the teammate, not
the prospect. Feed the parser an optional team roster (INTAKE_TEAM_ROSTER) via a
build_system(roster) outreach frame: roster names/initials are the people doing
outreach and are never extracted; the other party is the investor/prospect. Same
framing on the revise leg. Unset roster = prior behavior.
Turn the bot from a bare nohup process (silently dies on a Spark reboot) into a docker-compose service. Dockerfile bundles backend/matrix_intake + the stdlib backend/ingest Spark client it reuses; .env is mounted read-only at runtime, never baked. The existing repo-root .dockerignore (shared with the s9pk build) already keeps data/ and .env out of context. Also adds a handoff doc for wiring a spark-control dashboard card in a later session.
Close the two locked post-deploy enhancements for the Matrix intake bot.
Fuzzy matching (server-side, ships in the s9pk): new find_intake_candidates in
server.py returns ranked deterministic near-matches (difflib name similarity +
token-set Jaccard, legal-suffix-aware, + email Levenshtein <= 2); GET
/api/intake/match now returns {match, candidates}. The bot surfaces a numbered
shortlist so a near-duplicate (Charlie/Charles, Acme Capital vs Acme Capital LLC,
a one-char email typo) is confirmed by a human instead of silently creating a
second investor. Exact match still auto-attaches; fuzzy candidates are never
auto-attached. The optional LLM-judge re-rank is deferred.
Conversational edits (bot-side, ships on the Spark): any in-thread reply that
isn't yes/no/edit field=value is treated as a natural-language revision and
re-run through local Qwen (parse.revise). Email integrity is preserved -- a
changed address must literally appear in the instruction; the model's email
field is structurally unreachable. No-op revisions re-prompt.
Docs/current-state brought current; 27/27 backend tests green.
Four bot-side UX fixes surfaced by the live smoke:
- Post a brief pointer in the main timeline (a reply to the user's message)
alongside the in-thread proposal card, so proposals aren't missed inside a
thread. Pointer only — approvals still happen in the thread, where the note
is visible (you can't make an informed yes/no without seeing it).
- A bare yes/no typed in the main timeline while a proposal is pending now
gets a "reply in the thread" redirect instead of "couldn't tell what to record."
- Clearer commit confirmations: "Created a new grid entry for X" vs
"Logged a note on X (existing grid entry)."
- Send a blank communication subject when a note is present so the grid's
one-line note summary shows the note text, not the "(Matrix)" label
(provenance stays in source="matrix_intake").
normalize()'s email regex matched non-@/non-space runs, so "Name <addr>"
(the most common contact format) yielded "<addr"; only trailing punctuation
was stripped, never leading. Tighten the regex to standard local@domain.tld
so the bare address is extracted from <…>, (…), and trailing-period forms.
Found via the live-deploy pre-flight. Add a regression test.
Also log two intake backlog items in ROADMAP: the scoped service-credential
auth path (deferred; bot uses a member login for now) and fuzzy match +
in-thread confirm (post-deploy).
New backend/matrix_intake/ runs as its own process (matrix-nio isolated from the
stdlib CRM): local-Qwen parse via Spark Control → in-thread human approval
(yes/edit/no) → write through the CRM's own log-communication endpoint, tagged
source=matrix_intake. Adds read-only GET /api/intake/match (returns grid row id,
no-duplicate contract); threads provenance through handle_log_fundraising_communication.
Reviewer-passed: pop-before-commit closes a double-approve race; edit-grammar fix.
Text-only v1; business-card photo (M3) deferred (no Spark vision model).
26/26 tests green; live Matrix smoke pending deploy.
Keysat