Types

Conflict

A detected contradiction between two memory units.

Conflict

A Conflict represents a detected contradiction between two memory units. The protocol treats conflicts as expected events, not errors. Conflicts are declared explicitly by agents at every level (via a contradicts relation), and detected automatically at every level — mechanically (key-value disagreement, no embeddings) starting at Level 1, semantically at Level 2+.

A conflict's shape also depends on level. At Level 1, a conflict is a stateless fact: two unit IDs and the fields they disagree on — useful because there is nothing yet to do with it except surface it. At Level 2+, once MERGE exists to act on it, a conflict becomes a stateful, ID-addressable resource with a resolution lifecycle.

Schema

{
  "id": "<string | null>",
  "type": "factual | interpretive | strategic | priority | null",
  "status": "detected | resolving | resolved | escalated | null",
  "unit_a": "<string>",
  "unit_b": "<string>",
  "conflicting_fields": ["<string>"],
  "description": "<string>",
  "detected_by": "explicit | mechanical | semantic | logical | temporal",
  "resolution": {
    "strategy": "<ConflictStrategy>",
    "winner_id": "<string | null>",
    "rationale": "<string>",
    "resolved_by": "<string>",
    "epoch_resolved": "<number>"
  }
}

Field Reference

Field	Type	Required	Description
`id`	string or null	Level 2+	Unique conflict identifier, auto-assigned by the Field. `null` at Level 1 — a stateless conflict has no identity to reference later.
`type`	string or null	Level 2+	`factual`, `interpretive`, `strategic`, or `priority`. Unset at Level 1.
`status`	string or null	Level 2+	Current state of the conflict. Unset at Level 1.
`unit_a`	string	✓	ID of the first conflicting memory unit.
`unit_b`	string	✓	ID of the second conflicting memory unit.
`conflicting_fields`	string	For `mechanical` detection	The shared fields the two units disagree on.
`description`	string	Level 2+	Human-readable description of the contradiction.
`detected_by`	string	✓	How the conflict was discovered.
`resolution`	object		Populated once the conflict is resolved (Level 2+ only — there is no resolution lifecycle at Level 1).

Required Fields by Level

Field	Level 1	Level 2+
`unit_a`, `unit_b`	REQUIRED	REQUIRED
`detected_by`	REQUIRED	REQUIRED
`conflicting_fields`	REQUIRED for `mechanical` detection	REQUIRED for `mechanical` detection
`description`	OPTIONAL	REQUIRED
`id`	—	REQUIRED
`type`	—	REQUIRED
`status`	—	REQUIRED
`resolution`	—	REQUIRED once resolved

At Level 1, a Conflict is a stateless fact recomputed on demand — it has no identity to reference later and no lifecycle, because MERGE, the only operation that would advance it through one, doesn't exist yet.

Conflict Types

Type	Description
`factual`	Two units make opposing factual claims (e.g., conflicting market size figures)
`interpretive`	Two units interpret the same evidence differently
`strategic`	Two units recommend conflicting courses of action
`priority`	Two units assign different priority to the same goal or task

Detection Methods

Method	Description	Required Level
`explicit`	An agent included a `contradicts` relation in a RECORD request	1+
`mechanical`	Two committed units share a scoping key and disagree on the value of another shared primitive field. No embeddings or configuration required.	1+
`semantic`	Embedding similarity detected opposing claims	2+
`logical`	A finding's assumptions contradict another finding's content	2+
`temporal`	A newer finding contradicts an established, higher-confidence finding	2+

Status Lifecycle

Applies from Level 2+, once MERGE exists to drive it. At Level 1, a Conflict has no status to transition.

Conflict Lifecycle

detected → resolving → resolved
detected → escalated → resolved (human operator resolves)
resolving → escalated (strategy fails or times out)

Status	Description
`detected`	Conflict found; no resolution initiated yet.
`resolving`	An agent or the Field has initiated resolution with a chosen strategy.
`resolved`	Successfully resolved. Both original units updated to `superseded` or one is `superseded`.
`escalated`	Automatic resolution not possible; flagged for human review.

Conflict Resolution Strategies

Strategy	Description	Required Level
`last_write_wins`	Most recent memory unit prevails	2+
`confidence_weighted`	Higher confidence score prevails	2+
`human_escalation`	Flag for human review	2+
`authority`	Specified agent/role has final say	3
`evidence_count`	Unit with more supporting evidence prevails	3
`synthesis`	Create a new unit reconciling both	3
`vote`	Multiple agents vote (requires quorum)	3

JSON Schema

{
  "$schema": "https://json-schema.org/draft/2020-12/schema",
  "$id": "https://akashikprotocol.com/schema/0.1.0/conflict.json",
  "type": "object",
  "required": ["unit_a", "unit_b", "detected_by"],
  "properties": {
    "id": { "type": ["string", "null"], "minLength": 1 },
    "type": { "type": ["string", "null"], "enum": ["factual", "interpretive", "strategic", "priority", null] },
    "status": { "type": ["string", "null"], "enum": ["detected", "resolving", "resolved", "escalated", null] },
    "unit_a": { "type": "string" },
    "unit_b": { "type": "string" },
    "conflicting_fields": { "type": "array", "items": { "type": "string" } },
    "description": { "type": "string" },
    "detected_by": { "type": "string", "enum": ["explicit", "mechanical", "semantic", "logical", "temporal"] },
    "resolution": {
      "type": "object",
      "properties": {
        "strategy": {
          "type": "string",
          "enum": ["last_write_wins", "confidence_weighted", "authority", "evidence_count", "synthesis", "human_escalation", "vote"]
        },
        "winner_id": { "type": ["string", "null"] },
        "rationale": { "type": "string" },
        "resolved_by": { "type": "string" },
        "epoch_resolved": { "type": "integer" }
      }
    }
  }
}

Scope

Controls what an agent receives during attunement.

Message Envelope

The wrapper structure for every Akashik Protocol message.