@mindrian_os/install 1.13.0-beta.11

package/references/import-config.md

@@ -0,0 +1,141 @@
+---
+icm_layer: 3
+purpose: Stable configuration for the Phase 80 vault import pipeline
+audience: vault-import.cjs, classifier, router, enricher
+editable: once per workspace
+---
+
+# Import Config -- Layer 3 Reference
+
+This file is the single source of truth for classifier thresholds, role
+detection keywords, and the frontmatter promotion map used by every stage
+of the vault import pipeline (`scripts/vault-import.cjs` + `lib/import/*`).
+
+It is Layer 3 in ICM terms: stable across runs, edited once per workspace,
+consumed by stages as read-only reference data. Do NOT write per-run data
+here -- per-run state lives in `imports/{id}/MANIFEST.json`.
+
+---
+
+## Confidence Thresholds
+
+Confidence scores are floats in the inclusive range [0.0, 1.0]. The three
+routing decisions are computed purely from the classifier-produced score.
+
+| Decision | Confidence range   | Routing behavior                              |
+|----------|--------------------|-----------------------------------------------|
+| AUTO     | confidence >= 0.75 | Route directly to the suggested section      |
+| SUGGEST  | 0.45 - 0.74        | Route to `inbox/suggested/{slug}/{slug}.md`  |
+| INBOX    | confidence < 0.45  | Route to `inbox/unclassified/{slug}/{slug}.md` |
+
+```json
+{
+  "auto_threshold": 0.75,
+  "suggest_threshold": 0.45
+}
+```
+
+Review-gate edits can override any row in `02-classify/output/classifications.md`.
+The router reads the edited table, not the raw classifier output.
+
+---
+
+## Role Buckets
+
+Person detection produces a role guess per detected person. The guess is
+heuristic: the first bucket whose keyword list matches any mention context
+wins. Unmatched people land in `unassigned`. Larry can reassign at the
+review gate.
+
+```yaml
+role_buckets:
+  core-team:
+    keywords:
+      - founder
+      - co-founder
+      - cofounder
+      - CEO
+      - CTO
+      - COO
+      - CFO
+      - head of
+      - VP
+      - our team
+      - we built
+      - team lead
+  consultants:
+    keywords:
+      - consultant
+      - contractor
+      - freelance
+      - hired
+      - retained
+      - engaged
+  advisors:
+    keywords:
+      - advisor
+      - mentor
+      - guidance
+      - advised us
+      - coach
+  investors:
+    keywords:
+      - investor
+      - VC
+      - fund
+      - angel
+      - seed round
+      - Series A
+      - Series B
+      - Series C
+      - cap table
+      - term sheet
+  board:
+    keywords:
+      - board member
+      - chairman
+      - director
+      - sits on the board
+      - board seat
+  unassigned:
+    keywords: []
+```
+
+Matching is case-insensitive substring. Order matters: core-team is evaluated
+first so an explicit "co-founder" wins over a stray "advised us" mention.
+
+---
+
+## Frontmatter Promotion Map
+
+When an imported artifact has source frontmatter, a known subset is promoted
+into the artifact folder's ROOM.md (Layer 0 identity). Unknown fields stay
+in the artifact file's frontmatter untouched.
+
+```yaml
+promote_to_room_md:
+  - title
+  - aliases
+  - tags
+  - date
+  - created
+  - modified
+  - author
+  - attendees
+```
+
+Field semantics:
+
+| Source key  | ROOM.md destination           | Notes                                      |
+|-------------|-------------------------------|--------------------------------------------|
+| title       | ROOM.md identity              | Falls back to filename stem if missing     |
+| aliases     | ROOM.md identity              | Array preserved                            |
+| tags        | ROOM.md metadata + tag index  | Merged with section defaults               |
+| date        | ROOM.md metadata              | ISO-normalized                             |
+| created     | ROOM.md metadata              | ISO-normalized                             |
+| modified    | ROOM.md metadata              | ISO-normalized                             |
+| author      | ROOM.md metadata + person feed| Triggers person detection tier 1           |
+| attendees   | ROOM.md metadata + person feed| Triggers person detection tier 1           |
+
+Any other frontmatter key is preserved verbatim on the artifact file and
+NOT copied to ROOM.md.

package/references/integrations/detection-patterns.md

@@ -0,0 +1,101 @@
+# Integration Detection Patterns
+
+Reference for `lib/core/integration-registry.cjs` -- documents how each integration is detected, when Larry offers it, and the rules for non-blocking behavior.
+
+---
+
+## Integration Catalog
+
+### Brain (Teaching Graph)
+
+| Field | Value |
+|-------|-------|
+| **Detection** | `MINDRIAN_BRAIN_KEY` env var (from `.env`, `~/.mindrian.env`, or system env) |
+| **Benefit** | Enhanced framework suggestions, grading calibration, cross-domain connections |
+| **Setup** | `/mos:setup brain` |
+
+**Trigger signals:** User mentions "suggest-next", "grade", "connections", "framework" while Brain is not connected.
+
+**Offer text:** "By the way -- I'd be sharper with my teaching graph connected. One command: `/mos:setup brain`"
+
+---
+
+### Velma (Transcription)
+
+| Field | Value |
+|-------|-------|
+| **Detection** | `MODULATE_API_KEY` env var |
+| **Benefit** | Audio transcription with speaker diarization and emotion detection |
+| **Setup** | `/mos:setup transcription` |
+
+**Trigger signals:** User mentions "audio", "transcript", "recording", "meeting recording" while Velma is not connected.
+
+**Offer text:** "I can transcribe that with speaker identification if you connect Velma -- `/mos:setup transcription`"
+
+---
+
+### Obsidian (Vault Sync)
+
+| Field | Value |
+|-------|-------|
+| **Detection** | `.obsidian/` directory found in workspace or up to 3 parent directories |
+| **Benefit** | Vault sync -- room sections as Obsidian pages with graph view |
+| **Setup** | `/mos:setup obsidian` (future) |
+
+**Trigger signals:** User mentions "notes", "vault", "obsidian", "knowledge base" while no vault detected.
+
+**Offer text:** "I noticed you might use Obsidian. Room sections could sync there -- that's on the roadmap."
+
+---
+
+### Notion (Workspace Sync)
+
+| Field | Value |
+|-------|-------|
+| **Detection** | `NOTION_API_KEY` env var OR `notion` in .mcp.json mcpServers |
+| **Benefit** | Workspace sync -- room intelligence in Notion pages |
+| **Setup** | `/mos:setup notion` (future) |
+
+**Trigger signals:** User mentions "notion", "workspace", "wiki" while Notion is not connected.
+
+**Offer text:** "Notion sync is coming. For now, `/mos:export` gets your room into any format."
+
+---
+
+### Meeting Source (Auto-Fetch)
+
+| Field | Value |
+|-------|-------|
+| **Detection** | MCP server keys containing `read-ai`, `vexa`, or `recall-ai` in .mcp.json |
+| **Benefit** | Auto-fetch latest meeting transcripts |
+| **Setup** | `/mos:setup meetings` |
+
+**Trigger signals:** User mentions "latest meeting", "auto-fetch", "meeting source" while no meeting source connected.
+
+**Offer text:** "I can auto-fetch meeting transcripts if you connect a source -- `/mos:setup meetings`"
+
+---
+
+## Non-Blocking Rules
+
+These rules ensure integration offers NEVER interrupt the user's workflow:
+
+1. **Never during active methodology** -- if `roomState.activeMethodology` is truthy, suppress ALL offers
+2. **Maximum ONE offer per conversation** -- not per turn, per entire conversation
+3. **Answer first, offer second** -- the user's question always gets a complete answer before any integration mention
+4. **No repeats** -- if the user ignores an offer, do not repeat it in the same session
+5. **No pricing mentions** -- offers describe capability, never cost
+6. **No force** -- offers are suggestions ("By the way..."), never requirements ("You need to...")
+7. **Respect opted-out** -- if a user explicitly declines, never offer that integration again in the session
+
+## Context Signal Priority
+
+When multiple integrations could be suggested, the registry returns only the first match in catalog order:
+
+1. Brain (highest -- enriches everything)
+2. Velma (audio-specific)
+3. Obsidian (note-taking context)
+4. Notion (workspace context)
+5. Meeting source (meeting context)
+
+This ensures only the most impactful suggestion surfaces.

package/references/meeting/artifact-template.md

@@ -0,0 +1,377 @@
+# Artifact Template -- Meeting-Filed Entries
+
+*Used by `file-meeting` to create wicked-problem-aware artifacts from classified transcript segments.*
+
+---
+
+## YAML Frontmatter Template
+
+Every artifact filed from a meeting transcript MUST include this frontmatter:
+
+```yaml
+---
+methodology: file-meeting
+created: {YYYY-MM-DD}
+source: transcript
+attribution:
+  speaker: {name}
+  role: {role from 12-type taxonomy}
+  profile_path: team/{role-plural}/{speaker-slug}
+  meeting_date: {YYYY-MM-DD}
+  meeting_id: {YYYY-MM-DD-meeting-slug}
+segment_type: {type from 6-type taxonomy}
+confidence: {0.0-1.0}
+meeting_date: {YYYY-MM-DD}
+meeting_name: {meeting identifier}
+room_section: {target section from 8-section vocabulary}
+assumptions:
+  - claim: "{extracted assumption}"
+    status: unvalidated
+    impacts: [{affected sections}]
+perspective: {speaker_role}
+cascade_sections: [{sections this insight may affect}]
+secondary_types: []
+rejection: null
+---
+```
+
+---
+
+## Field Definitions
+
+### Core Fields
+
+| Field | Type | Required | Description |
+|-------|------|----------|-------------|
+| methodology | string | YES | Always `file-meeting` |
+| created | date | YES | Date the artifact was filed (YYYY-MM-DD) |
+| source | string | YES | `transcript` for text imports, `velma` for Velma-transcribed audio |
+| segment_type | enum | YES | One of the 6 types from segment-classification.md |
+| confidence | float | YES | 0.0-1.0 classification confidence |
+| meeting_date | date | YES | When the meeting occurred |
+| meeting_name | string | YES | Human-readable meeting identifier |
+| room_section | string | YES | Target room section for filing |
+
+### Attribution Block
+
+The `attribution:` block replaces the Phase 6 flat `speaker:` and `speaker_role:` fields. It provides a complete provenance chain linking each artifact to its speaker and meeting.
+
+| Field | Type | Required | Description |
+|-------|------|----------|-------------|
+| attribution.speaker | string | YES | Full name of the speaker |
+| attribution.role | enum | YES | One of the 12 roles from section-mapping.md |
+| attribution.profile_path | string | YES | Relative path to speaker profile (e.g., team/mentors/lawrence-aronhime) |
+| attribution.meeting_date | date | YES | When the meeting occurred (duplicated for provenance chain) |
+| attribution.meeting_id | string | YES | YYYY-MM-DD-{meeting-slug} unique meeting identifier |
+
+**Backward Compatibility:** compute-team and other scanners must handle BOTH flat `speaker:` (Phase 6 legacy) and nested `attribution.speaker:` (Phase 7+). Grep for `^speaker:` OR `^  speaker:` (indented under attribution:).
+
+### Wicked Problem Fields
+
+| Field | Type | Required | Description |
+|-------|------|----------|-------------|
+| assumptions | list | YES | Extracted assumptions with validity tracking |
+| perspective | string | YES | Matches speaker role -- frames the viewpoint |
+| cascade_sections | list | YES | Sections this artifact may impact beyond its target |
+
+### Optional Fields
+
+| Field | Type | Required | Description |
+|-------|------|----------|-------------|
+| secondary_types | list | NO | Additional segment types detected (multi-type segments) |
+| rejection | object | NO | Populated if user rejected a filing suggestion |
+| emotions | list | NO | Velma-detected emotions (source: velma only) |
+| velma_confidence | float | NO | Velma's transcription confidence (source: velma only) |
+
+---
+
+## Assumptions Field
+
+Every segment implies assumptions. Larry extracts them during filing.
+
+### Structure
+
+```yaml
+assumptions:
+  - claim: "TAM is $190M"
+    status: unvalidated
+    impacts: [financial-model, market-analysis]
+  - claim: "Enterprise sales cycles are 18 months"
+    status: unvalidated
+    impacts: [business-model, financial-model]
+```
+
+### Status Values
+
+| Status | Meaning |
+|--------|---------|
+| unvalidated | Default -- claim extracted but not verified |
+| validated | User or data confirmed the claim |
+| invalidated | Contradicted by newer information |
+| stale | Older than 90 days without revalidation |
+
+### Fallback
+
+When Larry cannot identify any assumptions in a segment:
+
+```yaml
+assumptions: none_detected
+```
+
+This is valid. Not every segment contains assumptions (e.g., pure action items like "Sarah will send the deck" imply no claims about the world).
+
+---
+
+## Cascade Sections Field
+
+Identifies which room sections BEYOND the target section may be affected by this artifact.
+
+### Cascade Detection Rules
+
+1. **Financial claim** -- always cascades to `financial-model`
+2. **Competitive mention** -- always cascades to `competitive-analysis`
+3. **Team capability claim** -- cascades to `team-execution`
+4. **Legal/regulatory mention** -- cascades to `legal-ip`
+5. **Problem reframing** -- cascades to `problem-definition` and potentially all sections
+6. **Market data** -- cascades to `market-analysis` and `business-model`
+
+### Example
+
+An advisor insight filed to `solution-design`:
+
+```yaml
+room_section: solution-design
+cascade_sections: [financial-model, team-execution]
+```
+
+The insight suggests a technical architecture change that will affect costs (financial-model) and require new hires (team-execution).
+
+---
+
+## Examples by Segment Type
+
+### decision
+
+```yaml
+---
+methodology: file-meeting
+created: 2026-03-15
+source: transcript
+attribution:
+  speaker: Sarah Chen
+  role: founder
+  profile_path: team/founders/sarah-chen
+  meeting_date: 2026-03-15
+  meeting_id: 2026-03-15-board-strategy-q1
+segment_type: decision
+confidence: 0.92
+meeting_date: 2026-03-15
+meeting_name: Board Strategy Session Q1
+room_section: solution-design
+assumptions:
+  - claim: "Enterprise segment is more viable than SMB"
+    status: unvalidated
+    impacts: [market-analysis, business-model, financial-model]
+perspective: founder
+cascade_sections: [market-analysis, business-model, financial-model]
+---
+
+We've decided to focus on the enterprise segment first. The SMB numbers don't support the unit economics yet. We'll pivot the product roadmap to enterprise features starting next sprint.
+```
+
+### action-item
+
+```yaml
+---
+methodology: file-meeting
+created: 2026-03-15
+source: transcript
+attribution:
+  speaker: David Park
+  role: team-member
+  profile_path: team/members/david-park
+  meeting_date: 2026-03-15
+  meeting_id: 2026-03-15-board-strategy-q1
+segment_type: action-item
+confidence: 0.88
+meeting_date: 2026-03-15
+meeting_name: Board Strategy Session Q1
+room_section: team-execution
+assumptions: none_detected
+perspective: team-member
+cascade_sections: []
+---
+
+David will prepare the enterprise pricing deck by next Friday and share it with the advisory board for feedback before the investor meeting.
+
+**Extracted action:**
+- owner: David Park
+- task: Prepare enterprise pricing deck
+- deadline: Next Friday (2026-03-22)
+- context: Needed before investor meeting
+```
+
+### insight
+
+```yaml
+---
+methodology: file-meeting
+created: 2026-03-15
+source: transcript
+attribution:
+  speaker: Lisa Wong
+  role: researcher
+  profile_path: team/researchers/lisa-wong
+  meeting_date: 2026-03-15
+  meeting_id: 2026-03-15-board-strategy-q1
+segment_type: insight
+confidence: 0.85
+meeting_date: 2026-03-15
+meeting_name: Board Strategy Session Q1
+room_section: market-analysis
+assumptions:
+  - claim: "Enterprise segment grew 34% last quarter"
+    status: unvalidated
+    impacts: [market-analysis, financial-model]
+perspective: researcher
+cascade_sections: [financial-model, business-model]
+---
+
+The enterprise segment grew 34% last quarter according to Gartner's latest report. That's almost double the SMB growth rate. The mid-market is actually shrinking.
+```
+
+### advice
+
+```yaml
+---
+methodology: file-meeting
+created: 2026-03-15
+source: transcript
+attribution:
+  speaker: Prof. Lawrence Aronhime
+  role: mentor
+  profile_path: team/mentors/lawrence-aronhime
+  meeting_date: 2026-03-15
+  meeting_id: 2026-03-15-mentoring-session
+segment_type: advice
+confidence: 0.90
+meeting_date: 2026-03-15
+meeting_name: Mentoring Session
+room_section: problem-definition
+assumptions:
+  - claim: "Problem framing determines solution space"
+    status: validated
+    impacts: [problem-definition, solution-design]
+perspective: mentor
+cascade_sections: [solution-design, market-analysis]
+---
+
+You should reframe the problem before jumping to solutions. Right now you're solving "how to sell to enterprises" but the real question might be "what problem do enterprises have that SMBs don't?" That reframing changes your entire approach.
+```
+
+### question
+
+```yaml
+---
+methodology: file-meeting
+created: 2026-03-15
+source: transcript
+attribution:
+  speaker: Michael Torres
+  role: investor
+  profile_path: team/investors/michael-torres
+  meeting_date: 2026-03-15
+  meeting_id: 2026-03-15-board-strategy-q1
+segment_type: question
+confidence: 0.87
+meeting_date: 2026-03-15
+meeting_name: Board Strategy Session Q1
+room_section: financial-model
+assumptions: none_detected
+perspective: investor
+cascade_sections: [business-model]
+---
+
+What's your current churn rate for the enterprise customers you already have? And how does that compare to your SMB churn?
+
+**Gap identified:** Enterprise churn rate data not available in current financial model.
+```
+
+### noise (flagged)
+
+```yaml
+---
+methodology: file-meeting
+created: 2026-03-15
+source: transcript
+attribution:
+  speaker: Sarah Chen
+  role: founder
+  profile_path: team/founders/sarah-chen
+  meeting_date: 2026-03-15
+  meeting_id: 2026-03-15-board-strategy-q1
+segment_type: noise
+confidence: 0.55
+meeting_date: 2026-03-15
+meeting_name: Board Strategy Session Q1
+room_section: competitive-analysis
+assumptions: none_detected
+perspective: founder
+cascade_sections: []
+---
+
+FLAGGED_NOISE: "Oh by the way, I ran into the CEO of Acme Corp at that conference last week."
+REASON: Contains competitor name "Acme Corp"
+ACTION: User review required -- may contain buried insight about competitive landscape.
+```
+
+---
+
+## Rejection Field
+
+When a user rejects a filing suggestion during confirm-then-file, the rejection reason becomes data:
+
+```yaml
+rejection:
+  action: rejected
+  reason: "This was just small talk, not a real insight"
+  timestamp: 2026-03-15T14:23:00Z
+```
+
+Or when deferred:
+
+```yaml
+rejection:
+  action: deferred
+  reason: "Need to verify the numbers first"
+  timestamp: 2026-03-15T14:23:00Z
+```
+
+Rejection reasons become graph data. "Why not" teaches the system as much as "yes."
+
+---
+
+## Source Distinction
+
+| Source Value | When Used | Extra Fields |
+|-------------|-----------|--------------|
+| transcript | Text transcript pasted or uploaded | -- |
+| velma | Audio transcribed via Velma API | emotions, velma_confidence |
+
+Velma-sourced artifacts include emotion metadata:
+
+```yaml
+source: velma
+emotions: [confident, analytical]
+velma_confidence: 0.94
+```
+
+---
+
+## Cross-Reference
+
+- Segment types: `references/meeting/segment-classification.md`
+- Speaker roles and routing: `references/meeting/section-mapping.md`
+- Transcript format detection: `references/meeting/transcript-patterns.md`
+- Summary format: `references/meeting/summary-template.md`
+- Speaker profiles: `references/meeting/speaker-profile-template.md`