H2oslabs Synnovatorzero Synnovator

Synnovator

File-based data engine for the Synnovator platform. All records are .md files with YAML frontmatter + Markdown body, stored under PROJECT_DIR/.synnovator/.

Quick Start

# Initialize data directory
uv run python .claude/skills/synnovator/scripts/engine.py --init

# CRUD operations
uv run python .claude/skills/synnovator/scripts/engine.py [--user USER_ID] COMMAND TYPE [OPTIONS]

Commands: create, read, update, delete

Data Model

7 content types stored as .md files in .synnovator/<type>/:

• event - Activities/competitions (YAML + Markdown body)
• post - User posts with tags (YAML + Markdown body)
• resource - File attachments (YAML frontmatter only)
• rule - Activity rules with scoring criteria (YAML + Markdown body)
• user - User accounts (YAML frontmatter only)
• group - Teams/groups (YAML frontmatter only)
• interaction - Likes, comments, ratings (YAML frontmatter only). Target info is stored in target_interaction relation, not on the interaction itself.

9 relationship types stored in .synnovator/relations/<type>/:

• event_rule - Activity-to-rule bindings
• event_post - Activity-to-post submissions
• event_group - Team activity registration
• event_event - Activity association (stage / track / prerequisite)
• post_post - Post references/embeds/replies
• post_resource - Post-to-attachment links
• group_user - Group membership (with approval workflow)
• user_user - User follow/block relationships
• target_interaction - Content-to-interaction bindings

See references/schema.md for complete field definitions.

Content CRUD

Create

engine.py [--user UID] create <type> --data '<json>' [--body 'markdown content']

Auto-generates: id, created_at, updated_at, deleted_at. Sets created_by from --user.

Read

engine.py read <type> --id <record_id>           # Single record
engine.py read <type> --filters '<json>'          # Filtered list
engine.py read <type> --include-deleted           # Include soft-deleted

Update

engine.py update <type> --id <record_id> --data '<json>'

For tags: "+tagname" appends, "-tagname" removes. Body: "_body": "new markdown".

Delete

engine.py delete <type> --id <record_id>          # Soft delete (default)
engine.py delete <type> --id <record_id> --hard    # Hard delete

Cascades: deleting a post removes its relations and soft-deletes linked interactions.

Relation CRUD

engine.py create <rel_type> --data '<json>'                        # Create
engine.py read <rel_type> --filters '<json>'                       # Read
engine.py update <rel_type> --filters '<json>' --data '<json>'     # Update
engine.py delete <rel_type> --filters '<json>'                     # Delete (hard)

Use _ or : separator: event_rule or event:rule.

Key Behaviors

• Soft delete: Sets deleted_at, record stays on disk. Default for content types.
• Hard delete: Removes file. Default for relations.
• Cache stats: like_count, comment_count, average_rating on posts are read-only — auto-recalculated when target_interaction relations are created/deleted. Manual updates to these fields are silently ignored.
• Two-step interactions: Creating an interaction requires: (1) create interaction for the record, (2) create target_interaction to link it to a target. Cache updates, duplicate-like checks, and target validation happen at step 2.
• Group approval: require_approval=true sets join status to pending; owner approves via UPDATE group_user.
• Uniqueness: Enforced for user (username), (email); like (user, target) at target_interaction creation; relation duplicates; user-per-event-per-group (a user can only belong to one group per event); user_user (source_user_id, target_user_id, relation_type); event_event (source_category_id, target_category_id).
• Self-reference prevention: user_user and event_event cannot reference the same entity as both source and target.
• Block enforcement: If user B blocks user A, A cannot follow B.
• Circular dependency detection: event_event with stage or prerequisite relation types prevents cycles (A→B→C→A).
• Prerequisite enforcement: Creating a event_group registration checks that all prerequisite events (via event_event prerequisite) are closed.
• Cascades: Deleting content cascades to relations and interactions per the schema spec. Notably:
  • Deleting a group cascades to both group_user and event_group relations.
  • Deleting a user cascades to group_user and user_user relations and soft-deletes all user interactions.
  • Deleting a event cascades to event_rule, event_post, event_group, and event_event relations and soft-deletes linked interactions.

Rule Enforcement

Rules linked to a event via event_rule are automatically enforced via a hook system. Rules support two definition styles: fixed fields (backward-compatible shorthand) and declarative checks (extensible condition-action pairs). All rules must pass (AND logic); any single violation rejects the operation.

Hook Points

Phases

• pre: Runs before the operation. on_fail: deny blocks the operation. on_fail: warn allows with warning. on_fail: flag marks but allows.
• post: Runs after a successful operation. Never blocks. Used for compute/award/notify actions.

Validation Chain

event_post:   event → event_rule → rule.checks[trigger=create_relation(event_post)]
group_user:      group → event_group → event → event_rule → rule.checks[trigger=create_relation(group_user)]
event_group:  event → event_event(prerequisite) + event_rule → rule.checks[trigger=create_relation(event_group)]
post.status:     post → event_post → event → event_rule → rule.checks[trigger=update_content(post.status)]
event.status: event → event_rule → rule.checks[trigger=update_content(event.status)]

Fixed Fields as Syntactic Sugar

Fixed fields on Rule (max_submissions, min_team_size, submission_format, etc.) are automatically expanded into equivalent checks entries by the engine. User-defined checks are appended after expanded entries.

Condition Types

Action Types (post phase only)

Behavior

• No rules linked → no constraints, operation proceeds normally.
• Rule field absent → that constraint is skipped (e.g., no submission_deadline means no time limit).
• Violation (pre) → operation rejected with error: Rule '<name>': <reason>.
• Violation (post) → action executes (flag/compute/award), does not rollback.
• Posts not linked to any event are unconstrained (rules only apply through event_post).

See docs/rule-engine.md for the full specification.

File Format

Each record is a .md file:

---
name: 2025 AI Hackathon
type: competition
status: published
id: cat_abc123
created_by: user_alice
created_at: '2025-01-01T00:00:00Z'
---

## Activity Description

Markdown content here.

Testing

Tests are organized as modular files under scripts/tests/, one per test-case spec in specs/testcases/.

# Run all tests
uv run python .claude/skills/synnovator/scripts/tests/run_all.py

# Run specific test files by number
uv run python .claude/skills/synnovator/scripts/tests/run_all.py 3 4 11

# Run a single test file directly
uv run python .claude/skills/synnovator/scripts/tests/test_11_user_journeys.py

Test File Map

Each file uses isolated data directories for parallel-safe execution. Shared fixtures are in scripts/tests/base.py.

See references/endpoints.md for detailed API examples. See references/schema.md for complete field and enum definitions.

Documentation Sync

This skill's data model is the implementation of the canonical specs in docs/ and specs/. When modifying the skill's schema, engine logic, or data model, must sync the following docs:

Checklist when updating the skill:

1. Update skill code (engine.py) and tests (scripts/tests/test_*.py)
2. Update skill reference docs (references/schema.md, references/endpoints.md)
3. Update canonical docs (docs/data-types.md, docs/relationships.md, docs/crud-operations.md, docs/user-journeys.md)
4. Run uv run python .claude/skills/synnovator/scripts/tests/run_all.py to verify

Converted and distributed by TomeVault — claim your Tome and manage your conversions.