Graphonomous Workflows

End-to-end recipes for common tasks. Pick the workflow that matches your situation.

Arguments

Workflow type: $ARGUMENTS

Cold Start (empty graph)

1. retrieve(action: "context", query: "project overview") — confirm empty

2. act(action: "manage_goal", operation: "create_goal", payload: {"title": "Orient on project", "horizon": "short"})

3. Explore code/docs → act(action: "store_node", ...) for each key finding

4. Link nodes with act(action: "store_edge", ...) where relationships improve retrieval

5. act(action: "manage_goal", operation: "set_progress", goal_id: "<id>", progress: 0.5)

6. consolidate(action: "run")

Warm Resume (existing knowledge)

1. act(action: "manage_goal", operation: "list_goals", filters: {"status": "active"})

2. route(action: "attention_survey", include_idle: false)

3. Pick top attention item or follow user request

4. retrieve(action: "context", query: "<current topic>")

5. Continue working with context, saving causal_context

Answer a Question

1. retrieve(action: "context", query: "<the question>") — save causal_context

2. Check topology.routing — if "deliberate", run route(action: "deliberate", query: "...")

3. Optionally retrieve(action: "coverage", query: "...") for quick act/learn/escalate pre-check

4. Answer using retrieved knowledge + current conversation

5. act(action: "store_node", node_type: "semantic", content: "...", source: "conversation")

6. If user confirms → learn(action: "from_outcome", status: "success", causal_node_ids: [...])

Explore a Codebase

1. Create parent goal: act(action: "manage_goal", operation: "create_goal", payload: {"title": "Explore <repo>"})

2. Create sub-goals per module (3–6), each with parent_goal_id

3. For each module: explore → act(action: "store_node", ...) → act(action: "store_edge", ...)

4. Use consolidate(action: "traverse", start_node_id: "<id>", max_depth: 2) to walk related knowledge

5. act(action: "manage_goal", operation: "link_nodes", ...) — link findings to sub-goals

6. act(action: "manage_goal", operation: "set_progress", ...) after each module

7. route(action: "review_goal", ...) on parent when all sub-goals done

Debug a Problem

1. retrieve(action: "context", query: "prior context on <issue>") — save causal_context

2. Isolate the issue using retrieved context

3. Test fix

4. act(action: "store_node", node_type: "episodic", content: "...")

5. learn(action: "from_outcome", status: "success"|"failure", causal_node_ids: [...]) with real IDs

6. Link to goal if applicable

Resolve Contradictions

1. Detect: retrieve returns nodes with contradicts edges, or routing: "deliberate"

2. Diagnose: route(action: "topology", query: "<topic>") — examine SCCs and fault-line edges

3. Deliberate: route(action: "deliberate", query: "<the contradiction>", write_back: true)

4. Verify: route(action: "topology", ...) again — confirm κ decreased

5. Store resolution as semantic node if not already written back

Ralph Loop (autonomous iteration)

1. route(action: "attention_survey") → pick top item

2. retrieve(action: "context", ...) for that goal — save causal_context

3. Act on it

4. learn(action: "from_outcome", ...) with real causal IDs

5. route(action: "review_goal", ...) → follow act/learn/escalate decision

6. Advance to next attention item

7. Repeat; consolidate(action: "run") every 4–5 cycles

Maintenance

1. consolidate(action: "stats") — check overall graph health

2. consolidate(action: "run")

3. act(action: "manage_goal", operation: "list_goals") — review all goals

4. Re-review any active goals near completion thresholds

5. Transition stale goals to suspended or abandoned with metadata

Teach a Domain

1. act(action: "manage_goal", operation: "create_goal", payload: {"title": "Learn <domain>"})

2. User provides info → act(action: "store_node", ...) for each fact/procedure

3. act(action: "store_edge", ...) to connect related concepts

4. act(action: "manage_goal", operation: "link_nodes", ...) — link all nodes to goal

5. Build out recursively until coverage review returns act

Multi-Agent Handoff

1. Prepare state snapshot: act(action: "manage_goal", operation: "list_goals")

2. For each active goal: act(action: "manage_goal", operation: "get_goal", goal_id: "<id>")

3. Include linked nodes, progress, outcomes, and coverage state

4. Recipient resumes with /graphonomous:bootstrap to load context

End of Session

1. Store pending knowledge: act(action: "store_node", ...)

2. Report outcomes: learn(action: "from_outcome", ...)

3. Update goal progress: act(action: "manage_goal", operation: "set_progress", ...)

4. Consolidate: consolidate(action: "run")