hypernexus-workflows

HyperNexus Workflows — Orchestration Templates

Composable multi-step workflows built from the 16 HyperNexus Desktop tools. Each workflow is a named pattern with clear entry criteria, tool sequence, and exit criteria.

Workflow 1: Memory-First Task Execution (MEM-FIRST)

When: Starting any multi-step task or feature build.

Step	Tool	Purpose
1	hypernexus_context_harvest	Pull relevant L2 context for the task topic
2	hypernexus_search_memory	Keyword search for prior decisions/gotchas
3	hypernexus_memory_vector_search	Conceptual recall for nuanced context
4	Execute the task	Do the work
5	hypernexus_add_fact	Store decisions, gotchas, non-obvious findings
6	hypernexus_record_observation	Record domain knowledge in the graph

Exit criteria: Facts stored for all non-obvious findings.

Workflow 2: Tool Discovery Pipeline (DISCOVER)

When: You need a capability you don't have locally.

Step	Tool	Purpose
1	hypernexus_tool_discover	Broad capability search by intent
2	hypernexus_mcp_search	Narrow search across 26k+ MCP tools
3	hypernexus_mcp_call	Route the call through the kernel
4	hypernexus_add_fact	Store what worked for future runs

Exit criteria: Tool found and called successfully.

Workflow 3: Knowledge Graph Building (GRAPH)

When: Learning a new codebase, system, or domain.

Step	Tool	Purpose
1	hypernexus_repomap	Get repository map with symbol summaries
2	hypernexus_code_search	Find key patterns and implementations
3	hypernexus_knowledge_graph (add_entities)	Register discovered entities
4	hypernexus_knowledge_graph (add_relations)	Wire entity relationships
5	hypernexus_record_observation	Attach observations to entities

Exit criteria: Entities + relations + observations persisted.

Workflow 4: Session Handoff (HANDOFF)

When: End of a work session or context switch.

Step	Tool	Purpose
1	hypernexus_session_context (get)	Read current goal/objective
2	hypernexus_add_fact	Store session summary
3	hypernexus_session_context (set)	Update goal for next session
4	hypernexus_record_observation	Record session learnings

Exit criteria: Session state persisted and goal updated.

Workflow 5: Production Runbook (OPS)

When: Deployments, incidents, health checks.

Step	Tool	Purpose
1	hypernexus_system_status	Check kernel + memory + MCP health
2	hypernexus_billing_status	Verify provider quotas
3	Execute ops action	Restart, deploy, fix
4	hypernexus_system_status	Verify health after action
5	hypernexus_add_fact	Record what was done and why

Exit criteria: Health verified post-action and change recorded.

Workflow 6: Deep Context Harvest (HARVEST)

When: Starting a complex multi-day task or resuming after a break.

Step	Tool	Purpose
1	hypernexus_context_harvest (depth=deep)	Pull deep L2 context
2	hypernexus_memory_list	Review recent facts
3	hypernexus_session_search	Find related past sessions
4	hypernexus_memory_vector_search	Conceptual recall
5	hypernexus_search_memory	Keyword recall

Exit criteria: Full context assembled before starting work.

Composition Rules

1. Always start with recall (Workflow 1 or 6) before any multi-step work.
2. Always end with persist (store fact + observation) for anything non-obvious.
3. Chain workflows: HARVEST → MEM-FIRST → HANDOFF for long sessions.
4. Use DISCOVER mid-flow when you hit a capability gap.
5. Use GRAPH when the task involves learning a system or domain.
6. Use OPS before and after any production change.

The Operating Loop (all workflows share this)

RECALL → DISCOVER (if needed) → EXECUTE → PERSIST → VERIFY