# Venture firm: an Omnigraph use case

A venture firm's deals, portfolio, people and investment theses, with every signal tied to the beliefs it supports or contradicts.

| | |
|---|---|
| Category | Investing |
| Graph ID | vcos |
| Sample data | Quito Capital, a fictional AI-infrastructure fund: 207 nodes, 460 edges |
| Schema | 17 node types, 67 edge types |
| Stored queries | 204 reads, 87 writes |
| Source | https://github.com/ModernRelay/omnigraph-cookbooks/tree/main/vc-os |

## Node types

### Organization

Organization is the universal entity-of-the-world node. It covers every business-like
entity: startups we engage with, organizations we observe in the market, LPs, peer VC
firms, hyperscaler acquirers, customers, regulators, accelerators. The `kind` enum
is the lens; `status` carries Quito's engagement state for `kind=startup` (nullable
otherwise). Quito itself is a Organization with `slug=org-quito, kind=vc-firm`.

- `slug: String` (key)
- `name: String`
- `brief: String?`
- `description: String?`
- `website: String?`
- `founded_year: I32?`
- `hq: String?`
- `kind: enum(startup, lp-institution, vc-firm, acquirer, customer, bank, regulator, association, accelerator, university, family-office, publisher, database, expert-network, other)`
- `status: enum(cold, watching, pipeline, evaluating, portfolio, exited, passed, observed)?`
- `reliability: enum(low, medium, high)?`
- `aum_usd_m: F64?`
- `fit_score: F64?`
- `tags: [String]?`
- `createdAt: DateTime`
- `updatedAt: DateTime`

### Person

Person is an individual human. Roles relative to Quito are NOT properties of the
person — they live on edges: `WorksAt org-quito` (team), `FounderOf co-x` (founder),
`LpInFund fund-iii` (individual LP), `RoleInDeal deal-y {role: expert}` (expert),
`DecisionMakerAt org-aws` (acquirer DM). This keeps the Person definition intrinsic.
Enrichment fields (prior_exits, years_operating, education, prior_organizations,
founder_score) support the "dynamically updating founder view" — typically populated
by a scraping/enrichment skill. `last_enriched_at` marks when that ran.

- `slug: String` (key)
- `name: String`
- `email: String?`
- `linkedin: String?`
- `location: String?`
- `bio: String?`
- `prior_exits: I32?`
- `years_operating: I32?`
- `education: String?`
- `prior_organizations: [String]?`
- `founder_score: F64?`
- `last_enriched_at: DateTime?`
- `tags: [String]?`
- `createdAt: DateTime`
- `updatedAt: DateTime`

### Deal

Deal is a funding event involving a Organization. A Deal can be Quito-engaged
(`FromFund` set, `LedByPerson` set, `our_check_size_usd_m` set) or externally
observed (none of those set, `outcome=observed`). PitchBook-imported rounds we
never engaged with use `outcome=observed`.
One Organization can have many Deals (multiple rounds). The CRM grain.

- `slug: String` (key)
- `name: String`
- `round_type: enum(pre-seed, seed, seed-extension, series-a, series-a-extension, series-b, series-c, series-d, series-d-plus, growth, bridge, secondary, other)?`
- `stage: enum(sourced, qualified, in-diligence, ic-ready, decided, closed, dead)`
- `outcome: enum(open, invested, passed, lost, withdrawn, observed)`
- `round_size_usd_m: F64?`
- `pre_money_valuation_usd_m: F64?`
- `post_money_valuation_usd_m: F64?`
- `our_check_size_usd_m: F64?`
- `our_ownership_delta_pct: F64?`
- `target_close_date: Date?`
- `date_announced: Date?`
- `date_closed: Date?`
- `brief: String?`
- `createdAt: DateTime`
- `updatedAt: DateTime`

### Fund

Fund — Quito's funds (or any fund tracked in the system).

- `slug: String` (key)
- `name: String`
- `vintage_year: I32?`
- `total_size_usd_m: F64?`
- `status: enum(raising, investing, harvesting, closed)?`
- `sector_focus: String?`
- `brief: String?`
- `createdAt: DateTime`
- `updatedAt: DateTime`

### Market

Market — sector/vertical/segment. A hub for sector-specialist Theses, Patterns,
and Lessons. Companies live in Markets via `OrganizationInMarket`.
`parent_market` is a free-text pointer (a slug) to a broader Market — promote to
an edge if hierarchical queries become load-bearing.

- `slug: String` (key)
- `name: String`
- `parent_market: String?`
- `brief: String?`
- `description: String?`
- `tags: [String]?`
- `createdAt: DateTime`
- `updatedAt: DateTime`

### Artifact

Artifact — raw source content with three storage modes:
  1. Inline text   — `body` populated (small text, transcripts, derived summaries)
  2. Native binary — `blob` populated (decks, audio, images, screenshots, PDFs)
  3. External live — `url` populated, content may drift (web pages, gmail messages)
  4. Pinned git    — `source=github` + `repo + path + commit_sha + branch + last_synced_at`
                      (partner-curated markdown that lives in a wiki repo;
                       `body` caches the text at `commit_sha`)
The `derivedFromArtifact` chain preserves provenance through transforms
(audio → transcript → summary → memo).
For the git-pinned mode (typically `kind=markdown, source=github`), the
design intent is:
  • The Artifact is the *current* pointer to a logical document
  • `commit_sha + last_synced_at` records last-known-good state
  • `body` is cached text at that commit (so queries don't network out)
  • Historical "what did the doc say on date X" goes through git history
    at the commit_sha captured by a graph branch snapshot of that date
  • Updates happen in place by an importer (cron or webhook); ArtifactDerivedFrom
    is reserved for content-transform lineage (transcript → summary), not
    for tracking every wiki edit.

- `slug: String` (key)
- `name: String`
- `kind: enum(email, ticket, chat-msg, meeting-note, transcript, deck, web-page, screenshot, doc, audio, image, summary, markdown)`
- `source: enum(email, chat, meeting-tool, web, doc-tool, crm, outbound, manual, derived, repo, other)`
- `source_app: String?`
- `title: String?`
- `body: String?`
- `blob: Blob?`
- `mime_type: String?`
- `url: String?`
- `repo: String?`
- `path: String?`
- `commit_sha: String?`
- `branch: String?`
- `last_synced_at: DateTime?`
- `thread_id: String?`
- `date: Date`
- `captured_at: DateTime`
- `createdAt: DateTime`
- `updatedAt: DateTime`

### Meeting

Meeting — a scheduled (or ad-hoc) event with structure that Artifact alone
can't carry: time, attendees, subject (deal/org/thesis), and outputs
(transcript, decisions, commitments). The transcript itself is an
`Artifact{kind=transcript|meeting-note}` linked via `ArtifactFromMeeting`;
the Meeting is what binds attendees, agenda, and outcomes around it.
JTBDs served (see README "Operational loops"):
  • IC brief — what's been said in prior meetings about this deal
  • Board prep — what was committed at the last board, due before the next
  • Founder/expert call history with a Person or Organization
  • Handoff — "what did Pawel discuss with Yuki?"
`kind=external-other` is the escape hatch for meetings that don't fit the
canonical VC cadences (dinners, conferences, ad-hoc 1:1s).

- `slug: String` (key)
- `name: String`
- `kind: enum(ic, board, partner-1on1, pipeline-review, portfolio-1on1, lp-update, founder-call, dd-call, expert-call, internal, external-other)`
- `scheduled_at: DateTime`
- `duration_min: I32?`
- `status: enum(scheduled, occurred, cancelled, rescheduled)`
- `location: String?`
- `summary: String?`
- `tags: [String]?`
- `createdAt: DateTime`
- `updatedAt: DateTime`

### Thesis

Thesis — a market- or pattern-level belief Quito has bet on. A Thesis is a
Pattern the firm has financialized. Theses become `contradicted` (not just
`retired`) when their underlying Assumptions accumulate contradicting Signals.

- `slug: String` (key)
- `name: String`
- `brief: String?`
- `description: String?`
- `status: enum(active, retired, contradicted)`
- `confidence: enum(low, medium, high)?`
- `formed_date: Date?`
- `tags: [String]?`
- `createdAt: DateTime`
- `updatedAt: DateTime`

### Assumption

Assumption — a falsifiable sub-belief grounding a Thesis or Decision.
`level` is the lens: is this a market belief, a founder belief, a product
belief, a competitive belief, a financial belief, or a strategic belief.

- `slug: String` (key)
- `name: String`
- `brief: String?`
- `description: String?`
- `level: enum(market, founder, product, competitive, financial, strategic)`
- `confidence: enum(low, medium, high)?`
- `createdAt: DateTime`
- `updatedAt: DateTime`

### Question

Question — an open uncertainty the firm needs to resolve before acting.
`priority` reflects urgency relative to upcoming decisions.

- `slug: String` (key)
- `name: String`
- `description: String?`
- `status: enum(open, partial, resolved)`
- `priority: enum(low, normal, high, urgent)?`
- `asked_at: Date?`
- `answer: String?`
- `createdAt: DateTime`
- `updatedAt: DateTime`

### Signal

Signal — a dated external fact. `date` is when the event happened in the world;
`stagingTimestamp` is when we recorded it (matches pharma-intel convention).
Signals are what move Beliefs: each Signal supports or contradicts an Assumption
or Thesis, or informs a Question.

- `slug: String` (key)
- `name: String`
- `kind: enum(discovery, launch, fundraise, exit, founder-event, market-move, competitive, customer, regulatory, team-change, portfolio-update, board-decision)`
- `brief: String?`
- `description: String?`
- `date: Date`
- `confidence: enum(low, medium, high)?`
- `impact: enum(low, normal, high)?`
- `stagingTimestamp: DateTime`
- `url: String?`
- `tags: [String]?`
- `createdAt: DateTime`
- `updatedAt: DateTime`

### Insight

Insight — a synthesized internal interpretation. Multi-agent IC simulation
uses two Insights with stance=bull and stance=bear against the same Deal,
each grounded in Signals via `InsightReliesOnSignal`. Founder-assessment
memos attach via `InsightAboutPerson`.
Note: `hypothesis` is NOT a valid kind here. An open prediction awaiting
evidence is a `Question`, which has the right lifecycle (open/partial/resolved)
and participates in the `DecisionNeedsAnswer` traversal.

- `slug: String` (key)
- `name: String`
- `kind: enum(memo, brief, observation, recap)`
- `stance: enum(bull, bear, neutral)?`
- `summary: String?`
- `body: String?`
- `createdAt: DateTime`
- `updatedAt: DateTime`

### Chunk

Chunk — IMPLEMENTATION DETAIL for hybrid search. Indexed slices of Artifact
content. Not an ontological commitment; if the storage stack changed, this
would change shape. v1 of this cookbook ships no Chunks in the seed. Prepare
raw Chunk JSONL with `omnigraph embed --input ... --output ... --spec ...
--reembed-all`, then load the generated file.

- `slug: String` (key)
- `text: String`
- `chunk_index: I32`
- `embedding: Vector(3072)`
- `createdAt: DateTime`

### Decision

Decision — a one-shot decision regarding a Deal, Thesis, or Organization. The
provenance chain (Decision → basedOnAssumption → contradictedBy → Signal
→ sourcedFromArtifact → publishedByOrganization) IS the audit trail. Combined
with snapshot-pinned reads, no separate audit log is needed.
Note: intros, follow-ups, schedule-a-second-meeting, and flag-at-next-board
are NOT Decisions — they're Commitments. A Decision is a one-shot act
(decided_at). A Commitment is an ongoing obligation (due_date).

- `slug: String` (key)
- `name: String`
- `kind: enum(invest, pass, follow-on, double-down, write-off, exit-plan, no-decision)`
- `rationale: String?`
- `decided_at: Date?`
- `createdAt: DateTime`
- `updatedAt: DateTime`

### Commitment

Commitment — a deferred action with a deadline and an assignee. Replaces the
"intro" or "follow-up" pseudo-decisions. Used for: intros to make, follow-ups
owed, board prep deliverables.

- `slug: String` (key)
- `name: String`
- `status: enum(open, in-progress, done, dropped, blocked)`
- `priority: enum(low, normal, high, urgent)?`
- `due_date: Date?`
- `completed_at: Date?`
- `detail: String?`
- `createdAt: DateTime`
- `updatedAt: DateTime`

### Pattern

Pattern — a recurring shape observed across many Deals, Signals, or Companies.
Patterns AGGREGATE (multiple subjects); Insights INTERPRET (typically one subject).
Patterns we have bet on become Theses. Patterns we have learned from become Lessons.

- `slug: String` (key)
- `name: String`
- `kind: enum(gtm, pricing, founder-archetype, market-timing, exit, failure-mode, tech-adoption, capital-structure, regulatory)`
- `brief: String?`
- `description: String?`
- `confidence: enum(low, medium, high)?`
- `createdAt: DateTime`
- `updatedAt: DateTime`

### Lesson

Lesson — a distilled OPERATIONAL rule meant to shape future behavior. Differs
from Insight (which is analytical/descriptive) in that Lessons are imperatives.
`kind=protocol` covers the runtime-rules use case (declarative behavior rules
the team wants agents to follow).
`status=tentative` means the Lesson lives on a review branch awaiting human
merge — the branch-based replacement for "tentative-learnings → review → promote".
`confidence` is epistemic (how strong is the evidence). They can differ: a high-
confidence Lesson can still be `tentative` if it just hasn't been merged yet.

- `slug: String` (key)
- `name: String`
- `kind: enum(protocol, rule-of-thumb, anti-pattern, runbook)`
- `body: String`
- `confidence: enum(low, medium, high)?`
- `status: enum(tentative, active, retired)`
- `createdAt: DateTime`
- `updatedAt: DateTime`

## Edge types

- `Deal -[ForOrganization]-> Organization`
- `Deal -[FromFund]-> Fund`
- `Deal -[LedByPerson]-> Person`
- `Deal -[RelevantThesis]-> Thesis`
- `Deal -[RelevantMarket]-> Market`
- `Deal -[LeadInvestor]-> Organization`
- `Deal -[ParticipantInvestor]-> Organization`
- `Person -[Knows]-> Person` Knows is stored bidirectionally by load discipline — the link_knows mutation creates both directions and the seed loads both (the inverse edge; @unique doesn't auto-create it). `@unique(@src, @dst)` now enforces pair-uniqueness as a true composite key, intra-batch + at merge; not declared here. See CLAUDE.md.
- `Person -[WorksAt]-> Organization` WorksAt is the canonical employment edge. Quito team members WorksAt org-quito (Organization kind=vc-firm). Founders may WorksAt the organization they founded (in addition to the FounderOf edge), or work at a different organization.
- `Person -[FounderOf]-> Organization`
- `Person -[BoardMemberAt]-> Organization`
- `Person -[RoleInDeal]-> Deal` RoleInDeal: any context-specific role a Person plays for a specific Deal. Use this for experts, customer references, co-investor leads, board candidates, venture partners advising on a deal, etc.
- `Person -[DecisionMakerAt]-> Organization`
- `Organization -[WouldAcquire]-> Organization` Exit-landscape edge — Organization → Organization (e.g., AWS would acquire Pinion Infer).
- `Organization -[LpInFund]-> Fund` LpInFund is now Organization → Fund. Institutional LPs are Companies with kind=lp-institution; individual LPs are family-office Companies (kind=family-office). Individuals working at LP institutions have WorksAt edges; the LP commitment lives at the Organization level.
- `Organization -[OrganizationInMarket]-> Market`
- `Signal -[SignalAboutOrganization]-> Organization` Signal classification — what the signal is about
- `Signal -[SignalAboutPerson]-> Person`
- `Signal -[SignalAboutMarket]-> Market`
- `Signal -[SignalAboutDeal]-> Deal`
- `Signal -[SupportsThesis]-> Thesis` Signal → Belief (the analytical loop)
- `Signal -[ContradictsThesis]-> Thesis`
- `Signal -[SupportsAssumption]-> Assumption`
- `Signal -[ContradictsAssumption]-> Assumption`
- `Signal -[InformsQuestion]-> Question`
- `Thesis -[ThesisReliesOnAssumption]-> Assumption` Thesis structure
- `Thesis -[ThesisAboutMarket]-> Market`
- `Thesis -[ThesisAboutOrganization]-> Organization`
- `Question -[QuestionAboutDeal]-> Deal` Question targeting
- `Insight -[InsightAboutDeal]-> Deal` Insight wiring
- `Insight -[InsightAboutOrganization]-> Organization`
- `Insight -[InsightAboutPerson]-> Person`
- `Insight -[InsightHighlightsPattern]-> Pattern`
- `Insight -[InsightReliesOnSignal]-> Signal`
- `Signal -[SignalSourcedFromArtifact]-> Artifact` Provenance (Artifact ↔ world)
- `Artifact -[PublishedByOrganization]-> Organization` Publishing-source organization (publisher / database / expert-network). When the org's `reliability` drops, walk this edge backward through SignalSourcedFromArtifact to flag every downstream Signal for revalidation.
- `Artifact -[ArtifactFromPerson]-> Person`
- `Artifact -[ArtifactAboutDeal]-> Deal`
- `Artifact -[ArtifactAboutOrganization]-> Organization`
- `Artifact -[ArtifactMentionsPerson]-> Person`
- `Artifact -[ArtifactDerivedFrom]-> Artifact`
- `Artifact -[ArtifactDocumentsThesis]-> Thesis` Documents-edges: a curated belief node (Thesis/Lesson/Pattern) is *backed* by a markdown artifact in the wiki repo. The structured node owns the relational shape (status, confidence, edges); the markdown owns the prose. Typically the linked Artifact has `kind=markdown, source=github` and the git-pinned fields populated. One-to-one is expected but not enforced — multiple drafts/revisions are possible during edit cycles.
- `Artifact -[ArtifactDocumentsLesson]-> Lesson`
- `Artifact -[ArtifactDocumentsPattern]-> Pattern`
- `Chunk -[ChunkOfArtifact]-> Artifact` Chunk for hybrid search
- `Decision -[DecisionRegardingDeal]-> Deal`
- `Decision -[DecisionRegardingThesis]-> Thesis`
- `Decision -[DecisionBasedOnAssumption]-> Assumption`
- `Decision -[DecisionNeedsAnswer]-> Question`
- `Decision -[DecisionByPerson]-> Person`
- `Commitment -[CommitmentForPerson]-> Person`
- `Commitment -[CommitmentRegardingDeal]-> Deal`
- `Commitment -[CommitmentAssignedTo]-> Person`
- `Commitment -[CommitmentFromArtifact]-> Artifact`
- `Pattern -[PatternAcrossDecision]-> Decision` Reflexive: Pattern + Lesson
- `Pattern -[PatternAcrossSignal]-> Signal`
- `Pattern -[PatternAcrossOrganization]-> Organization`
- `Lesson -[LessonDistilledFromPattern]-> Pattern`
- `Lesson -[LessonAppliesToMarket]-> Market`
- `Meeting -[MeetingAboutDeal]-> Deal` ═══════════════════════════════════════════════════════════════════ EDGES — Meetings (operational layer) ═══════════════════════════════════════════════════════════════════ Meeting subject: what the meeting was about. A meeting can be about multiple subjects (e.g. a partner 1:1 that discussed 3 deals); load all.
- `Meeting -[MeetingAboutOrganization]-> Organization`
- `Meeting -[MeetingAboutThesis]-> Thesis`
- `Meeting -[MeetingAboutMarket]-> Market`
- `Meeting -[MeetingAttendedBy]-> Person` Meeting attendance. The `role` discriminates host/attendee/optional/declined so the same edge type covers invites and acceptances. Bind the edge in a query (`$meeting $attendance:meetingAttendedBy $person`) to project or filter on `$attendance.role`.
- `Artifact -[ArtifactFromMeeting]-> Meeting` Meeting outputs. Direction follows existing convention (the dependent entity points back to the source), matching `ArtifactFromPerson`, `CommitmentFromArtifact`, `SignalSourcedFromArtifact`.
- `Decision -[DecisionFromMeeting]-> Meeting`
- `Commitment -[CommitmentFromMeeting]-> Meeting`

Full schema: https://github.com/ModernRelay/omnigraph-cookbooks/blob/main/vc-os/schema.pg

## Questions it answers

### Which active theses has new evidence contradicted?

All active theses with at least one contradicting signal

```graphql
// All active theses with at least one contradicting signal
query contradicted_active_theses() {
    match {
        $t: Thesis { status: "active" }
        $s contradictsThesis $t
    }
    return { $t.slug, $t.name, $t.confidence, $s.slug, $s.name, $s.date, $s.brief }
    order { $s.date desc }
}
```

### Before investment committee, what contradicts the assumptions behind this deal?

```graphql
query pre_ic_brief_evidence($slug: String) {
    match {
        $d: Deal { slug: $slug }
        $d relevantThesis $t
        $t thesisReliesOnAssumption $a
        $s contradictsAssumption $a
    }
    return { $a.slug, $a.name, $s.slug, $s.name, $s.kind, $s.date, $s.brief }
    order { $s.date desc }
}
```

### Who on the team can get an introduction to this founder?

2-hop: team member → bridge person → founder. Constrains the target to a Person with at least one FounderOf edge so non-founders don't slip through. NOTE: This pattern *requires* a bridge — if a teammate knows the founder directly, no row is returned. To cover the full JTBD "who can intro me?", also call `intro_direct_to_founder` (1-hop) and union the results.

```graphql
// 2-hop: team member → bridge person → founder. Constrains the target to a
// Person with at least one FounderOf edge so non-founders don't slip through.
//
// NOTE: This pattern *requires* a bridge — if a teammate knows the founder
// directly, no row is returned. To cover the full JTBD "who can intro me?",
// also call `intro_direct_to_founder` (1-hop) and union the results.
query intro_path_to_founder($slug: String) {
    match {
        $founder: Person { slug: $slug }
        $founder founderOf $f_organization
        $bridge knows $founder
        $teammate: Person
        $teammate knows $bridge
        $teammate worksAt $quito
        $quito: Organization { slug: "org-quito" }
    }
    return { $teammate.slug, $teammate.name, $bridge.slug, $bridge.name, $f_organization.slug }
}
```

### What has happened at a portfolio company since the last board meeting?

Everything bearing on the next board meeting for this portco: recent signals, open questions, open commitments.

```graphql
// Everything bearing on the next board meeting for this portco:
// recent signals, open questions, open commitments.
query board_prep_pack($slug: String) {
    match {
        $c: Organization { slug: $slug }
        $s signalAboutOrganization $c
    }
    return { $s.slug, $s.name, $s.kind, $s.date, $s.brief }
    order { $s.date desc }
}
```

## Stored queries

### queries/beliefs.gq

- `get_thesis`
- `theses_active`
- `theses_contradicted`
- `theses_retired`
- `thesis_assumptions`
- `thesis_supporting_signals`
- `thesis_contradicting_signals`
- `thesis_market`
- `thesis_organizations`
- `thesis_deals`
- `thesis_contradictions_via_assumptions`: Pre-board / pre-IC scope: this thesis is contradicted by which signals?
- `get_assumption`
- `assumptions_by_level`
- `assumption_supports`
- `assumption_contradictions`
- `assumption_theses`
- `assumption_decisions`
- `get_question`
- `open_questions`
- `questions_by_status`
- `question_signals`
- `question_decisions`
- `question_deal`

### queries/deals.gq

- `get_deal`
- `deals_open`
- `deals_by_stage`
- `deals_by_outcome`
- `search_deals`
- `deal_organization`
- `deal_fund`
- `deal_lead_partner`
- `deal_thesis`
- `deal_market`
- `deal_lead_investor`
- `deal_participants`
- `deal_people`
- `deal_signals`
- `deal_artifacts`
- `deal_insights`
- `deal_open_questions`
- `deal_decisions`
- `deal_commitments`
- `debate_stances`: Multi-agent IC simulation — bull and bear stances grounded in real Signals.

### queries/decisions.gq

- `get_decision`
- `decisions_recent`
- `decisions_by_kind`
- `decision_deal`
- `decision_thesis`
- `decision_assumptions`
- `decision_questions`
- `decision_by_person`
- `get_commitment`
- `commitments_open`
- `commitments_by_status`
- `commitment_for_person`
- `commitment_assigned_to`
- `commitment_deal`
- `commitment_artifact`

### queries/meetings.gq

- `get_meeting`
- `meetings_recent`
- `meetings_upcoming`
- `meetings_by_kind`
- `meetings_by_status`
- `search_meetings`
- `meeting_deal`
- `meeting_organization`
- `meeting_thesis`
- `meeting_market`
- `meeting_attendees`
- `meetings_attended_by`
- `meeting_artifacts`
- `meeting_decisions`
- `meeting_commitments`
- `meetings_with_organization`: All meetings explicitly tagged to this organization (direct subject only; for deal-mediated meetings use `meetings_via_deal_for_organization`).
- `meetings_via_deal_for_organization`: All meetings about deals attached to this organization. Pair with `meetings_with_organization` for the full meeting history.
- `meetings_with_deal`
- `meetings_with_thesis`
- `board_meetings_for_organization`: All occurred board meetings for an organization, time-sorted (most recent first). Agent or UI takes the head row as "last board meeting."
- `upcoming_board_meetings_for_organization`: All scheduled (not yet occurred) board meetings for an organization. Head row is the next board.
- `ic_prep_meeting_history`: All occurred meetings for a deal, with kind + summary. Use as the "what we've discussed so far" panel in pre-IC briefs.
- `ic_prep_open_commitments`: Open commitments tied to meetings that have already occurred for this deal — "what's still owed before we IC." The $m.status="occurred" filter prevents commitments recorded ahead of a *future* meeting from leaking into the pre-IC view; only obligations from past conversations count as outstanding work.

### queries/mutations.gq

- `add_organization` (write)
- `add_person` (write)
- `add_market` (write)
- `add_deal` (write)
- `add_fund` (write)
- `add_artifact` (write)
- `add_meeting` (write)
- `add_markdown_artifact` (write): Git-pinned markdown artifact. `kind` defaults to "markdown" and `source` to "github" semantically; pass them in for forward-compat with other VCS or doc kinds. `body` carries the cached text at `commit_sha`; importer/cron updates both in lockstep.
- `add_thesis` (write)
- `add_assumption` (write)
- `add_question` (write)
- `add_signal` (write)
- `add_insight` (write)
- `add_insight_with_stance` (write): IC-debate variant: same as add_insight but also sets stance + summary at create time, so the resulting Insight participates in debate-stances reads. Multi-agent IC simulation (bull + bear in parallel) uses this.
- `add_publisher_organization` (write): Convenience helper for publisher / database / expert-network orgs with a reliability rating. Use the generic `add_organization` for everything else.
- `add_chunk` (write)
- `add_decision` (write)
- `add_commitment` (write)
- `add_pattern` (write)
- `add_lesson` (write)
- `link_deal_organization` (write)
- `link_deal_fund` (write)
- `link_deal_lead` (write)
- `link_deal_thesis` (write)
- `link_deal_market` (write)
- `link_deal_lead_investor` (write)
- `link_deal_participant` (write)
- `link_knows` (write): Knows is symmetric — this mutation inserts both directions in one call so callers can't accidentally create a one-way relationship. Knows does not declare @unique(src, dst), so this mutation remains the convention.
- `link_works_at` (write)
- `link_founder_of` (write)
- `link_board_member_at` (write)
- `link_role_in_deal` (write)
- `link_decision_maker_at` (write)
- `link_would_acquire` (write)
- `link_lp_in_fund` (write)
- `link_organization_in_market` (write)
- `link_signal_about_organization` (write)
- `link_signal_about_person` (write)
- `link_signal_about_market` (write)
- `link_signal_about_deal` (write)
- `link_supports_thesis` (write)
- `link_contradicts_thesis` (write)
- `link_supports_assumption` (write)
- `link_contradicts_assumption` (write)
- `link_informs_question` (write)
- `link_thesis_relies_on_assumption` (write)
- `link_thesis_about_market` (write)
- `link_thesis_about_organization` (write)
- `link_question_about_deal` (write)
- `link_insight_about_deal` (write)
- `link_insight_about_organization` (write)
- `link_insight_about_person` (write)
- `link_insight_highlights_pattern` (write)
- `link_insight_relies_on_signal` (write)
- `link_signal_sourced_from_artifact` (write)
- `link_published_by_organization` (write)
- `link_artifact_from_person` (write)
- `link_artifact_about_deal` (write)
- `link_artifact_about_organization` (write)
- `link_artifact_mentions_person` (write)
- `link_artifact_derived_from` (write)
- `link_artifact_documents_thesis` (write)
- `link_artifact_documents_lesson` (write)
- `link_artifact_documents_pattern` (write)
- `link_chunk_of_artifact` (write)
- `link_decision_regarding_deal` (write)
- `link_decision_regarding_thesis` (write)
- `link_decision_based_on_assumption` (write)
- `link_decision_needs_answer` (write)
- `link_decision_by_person` (write)
- `link_commitment_for_person` (write)
- `link_commitment_regarding_deal` (write)
- `link_commitment_assigned_to` (write)
- `link_commitment_from_artifact` (write)
- `link_pattern_across_decision` (write)
- `link_pattern_across_signal` (write)
- `link_pattern_across_organization` (write)
- `link_lesson_distilled_from` (write)
- `link_lesson_applies_to_market` (write)
- `link_meeting_about_deal` (write)
- `link_meeting_about_organization` (write)
- `link_meeting_about_thesis` (write)
- `link_meeting_about_market` (write)
- `link_meeting_attended_by` (write)
- `link_artifact_from_meeting` (write)
- `link_decision_from_meeting` (write)
- `link_commitment_from_meeting` (write)

### queries/organizations.gq

- `get_organization`
- `organizations_by_kind`
- `organizations_by_status`
- `portfolio_organizations`
- `pipeline_organizations`
- `observed_organizations`
- `search_organizations`
- `high_fit_pipeline`
- `organization_deals`
- `organization_founders`
- `organization_board`
- `organization_signals`
- `organization_artifacts`
- `organization_insights`
- `organization_markets`
- `organization_theses`
- `co_investor_leads`: VC firms that led the rounds we invested in.
- `co_investor_participants`: VC firms that participated in the rounds we invested in.
- `organization_co_investor_leads`: Given an Organization we invested in, which other VCs LED the round. Split into two queries because the query language has no OR primitive — callers (or aliases) union the leads + participants for the full co-investor view.
- `organization_co_investor_participants`: Given an Organization we invested in, which other VCs PARTICIPATED in the round.
- `vc_firm_coinvestments`: Given a VC firm, the deals where they participated alongside us.
- `exit_landscape`: Companies that could acquire this one.
- `exit_landscape_decision_makers`: Decision-makers at those acquirers.
- `get_market`
- `markets_all`
- `market_organizations`
- `market_theses`
- `market_signals`
- `market_lessons`

### queries/patterns.gq

- `get_pattern`
- `patterns_all`
- `patterns_by_kind`
- `pattern_signals`
- `pattern_decisions`
- `pattern_organizations`
- `pattern_insights`
- `get_lesson`
- `lessons_active`
- `lessons_tentative`
- `lessons_by_kind`
- `lesson_pattern`
- `lesson_markets`
- `lessons_for_market`: Sector-specialist runbook: all lessons applying to a given market
- `protocols`: Active protocols across the firm

### queries/people.gq

- `get_person`
- `search_people`
- `founders_enriched`: Founders with strong enrichment data
- `team`: Quito team — anyone WorksAt org-quito.
- `founders`: Anyone who has founded at least one Organization.
- `lp_contacts`: Anyone working at an LP-institution Organization.
- `acquirer_decision_makers`: Anyone working at an acquirer Organization (corp dev, M&A leads).
- `deal_role_participants`: Persons playing any role on a deal (founder, customer-ref, expert, co-investor, …). Bind the traversed edge so its role can be projected and filtered.
- `person_works_at`
- `person_founder_of`
- `person_board_seats`
- `person_role_in_deal`
- `person_decision_maker_at`
- `intro_path_to_founder`: 2-hop: team member → bridge person → founder. Constrains the target to a Person with at least one FounderOf edge so non-founders don't slip through. NOTE: This pattern *requires* a bridge — if a teammate knows the founder directly, no row is returned. To cover the full JTBD "who can intro me?", also call `intro_direct_to_founder` (1-hop) and union the results.
- `intro_direct_to_founder`: 1-hop direct: which team members directly know this founder. Companion to `intro_path_to_founder` — together they cover both "warm intro" and "I know them already" cases. Constrains the target to be a founder so non-founders don't slip through.
- `direct_team_knowers`: 1-hop: which team members directly know this person.
- `close_network`: Full close network for one person (one-hop).
- `person_decisions`: Decisions this person made.
- `person_open_commitments`: Open commitments assigned to this person.
- `person_commitments_for`: What's owed TO this person (intros to make, follow-ups owed).
- `person_signals`: Recent signals about a person (founder events, team changes).
- `person_insights`: Founder-assessment Insights attached to a Person.

### queries/portfolio.gq

- `portfolio_book`: The book: every portfolio organization.
- `fund_portfolio`: Portfolio organizations in a fund (via Deals → Fund)
- `portco_rounds`: All deals (rounds) for a portfolio organization — gives position history
- `portfolio_updates_recent`: Operating updates across the portfolio, time-sorted (most recent first). There is no window-function form of "latest per organization" here; the agent or UI can take the first row per organization from this ordered stream.
- `portfolio_recent_signals`: All signals about portfolio organizations, most recent first
- `portfolio_open_questions`: Open questions across all portfolio organizations' deals
- `portfolio_open_commitments`: Open commitments on portfolio deals
- `board_prep_pack`: Everything bearing on the next board meeting for this portco: recent signals, open questions, open commitments.
- `board_prep_open_questions`
- `board_prep_commitments`
- `pre_ic_brief_organization`: A single deal's full belief + evidence + commentary picture
- `pre_ic_brief_thesis`
- `pre_ic_brief_evidence`
- `pre_ic_brief_questions`
- `pre_ic_brief_debate`
- `reserve_pressure`: Which portfolio organizations in this fund have open follow-on-trigger questions?
- `contradicted_active_theses`: All active theses with at least one contradicting signal

### queries/signals.gq

- `get_signal`
- `recent_signals`
- `signals_by_kind`
- `search_signals`
- `signal_organization`
- `signal_person`
- `signal_market`
- `signal_deal`
- `signal_artifact`
- `signal_supports_assumptions`
- `signal_contradicts_assumptions`
- `signal_supports_theses`
- `signal_contradicts_theses`
- `signal_informs_questions`
- `signal_portfolio_impact`: A new signal arrives — which committed portfolio Decisions does it threaten? Walks: Signal contradicts → Assumption ← basedOn ← Decision → regarding → Deal → forOrganization (status=portfolio)
- `signal_engagement_impact`: Broader than portfolio: any *engaged* organization (watching, pipeline, evaluating, or portfolio) whose active Deal Decisions rest on an assumption this signal contradicts. Use this when a signal lands and you want to know every live deal it touches — not just closed investments.
- `recent_contradictions`: Recent contradictions across the firm

### queries/sources.gq

- `get_source`
- `sources_by_kind`: All sources: publishers (TechCrunch, etc.), databases (PitchBook), expert networks (Tegus). The query language has no enum-IN filter or OR primitive, so "all sources" is composed by the caller from the three by-kind queries below. `sources_by_reliability` filters across all three kinds in one query because every source-type org has a reliability set.
- `publishers`
- `databases`
- `expert_networks`
- `sources_by_reliability`
- `source_artifacts`: Artifacts published by this source.
- `source_downstream_signals`: Signals whose artifact was published by this source. If reliability drops, these are the signals to revalidate.
- `artifact_source`: The reverse: given an Artifact, what organization published it.

### queries/wiki.gq

- `get_markdown_artifact`: Get a single markdown Artifact by slug. Filtered to kind=markdown so callers passing a non-markdown slug get 0 rows back rather than a partially-null shape that pretends to be one.
- `markdown_artifacts_recent`: All repo-stored artifacts, most-recently-synced first. `source=repo` covers any git-backed content; narrow to a specific provider via `source_app`.
- `markdown_artifacts_in_repo`: All repo-stored artifacts in a specific repo (so a fund can run multiple wikis).
- `markdown_artifacts_only`: Markdown artifacts specifically — filters out other future repo-sourced kinds.
- `thesis_doc`
- `lesson_doc`
- `pattern_doc`
- `markdown_documents_thesis`
- `markdown_documents_lesson`
- `markdown_documents_pattern`
- `theses_active_for_doc_review`
- `theses_with_docs`
- `lessons_active_for_doc_review`
- `lessons_with_docs`

## Run it

```bash
git clone https://github.com/ModernRelay/omnigraph-cookbooks
cd omnigraph-cookbooks
export COOKBOOK=$PWD/vc-os
omnigraph cluster validate --config "$COOKBOOK"
omnigraph cluster apply --target "$COOKBOOK" --config "$COOKBOOK" --as act-admin --json
```

The full local setup (server, sample data, queries): https://github.com/ModernRelay/omnigraph-cookbooks#local-setup

Source: https://www.omnigraph.dev/use-cases/vc-os
