{
  "$schema": "https://json-schema.org/draft/2020-12/schema",
  "$id": "https://sediment.so/schemas/derived-artifacts/attributed-completion/v2.json",
  "title": "AttributedCompletion",
  "description": "One canonical attributed completion.",
  "type": "object",
  "properties": {
    "org_id": {
      "type": "string",
      "description": "The deployment's tenant id, normalized. Bound to the server."
    },
    "session_id": {
      "type": "string",
      "description": "The coding-agent session this belongs to (ADR 0002)."
    },
    "inference_call_id": {
      "type": "string",
      "description": "Id of the inference call this record belongs to."
    },
    "repo": {
      "anyOf": [
        {
          "type": "string"
        },
        {
          "type": "null"
        }
      ],
      "description": "The attributed repo. Null on an abandonment-evidence row; Sediment does not guess a repo for work that reached no commit."
    },
    "commit_sha": {
      "anyOf": [
        {
          "type": "string"
        },
        {
          "type": "null"
        }
      ],
      "description": "The attributed commit. Null on an abandonment-evidence row."
    },
    "file_path": {
      "anyOf": [
        {
          "type": "string"
        },
        {
          "type": "null"
        }
      ],
      "description": "The changed file the attribution was keyed on. Null on an abandonment-evidence row."
    },
    "similarity_score": {
      "anyOf": [
        {
          "type": "number"
        },
        {
          "type": "null"
        }
      ],
      "description": "The attribution's similarity score, carried through so a projection can discount on it without re-deriving. Null on an abandonment-evidence row."
    },
    "attribution_source": {
      "anyOf": [
        {
          "enum": [
            "git_notes",
            "jaccard"
          ],
          "type": "string"
        },
        {
          "type": "null"
        }
      ],
      "description": "The attribution's `git_notes` or `jaccard` source. Null on an abandonment-evidence row."
    },
    "decisions": {
      "type": "array",
      "items": {
        "$ref": "#/$defs/DeveloperDecision"
      },
      "description": "The developer decisions recorded against this completion."
    },
    "ci_outcomes": {
      "type": "array",
      "items": {
        "$ref": "#/$defs/CIOutcome"
      },
      "description": "The CI runs recorded for this commit."
    },
    "provenance": {
      "$ref": "#/$defs/Provenance",
      "description": "The structured provenance for this artifact."
    },
    "split": {
      "enum": [
        "train",
        "eval"
      ],
      "type": "string",
      "description": "Which side of the deterministic eval split this row is on, hashed on `session_id` so a session never straddles both."
    },
    "abandonment": {
      "anyOf": [
        {
          "$ref": "#/$defs/SessionAbandonment"
        },
        {
          "type": "null"
        }
      ],
      "description": "The session-level abandonment evidence for an explicit-accept negative. Null on an attribution-evidence row. Exactly one evidence variant must be present."
    }
  },
  "additionalProperties": false,
  "required": [
    "org_id",
    "session_id",
    "inference_call_id",
    "repo",
    "commit_sha",
    "file_path",
    "similarity_score",
    "attribution_source",
    "decisions",
    "ci_outcomes",
    "provenance",
    "split",
    "abandonment"
  ],
  "$defs": {
    "DeveloperDecision": {
      "type": "object",
      "properties": {
        "decision_id": {
          "type": "string",
          "description": "Stable id for this decision."
        },
        "org_id": {
          "type": "string",
          "description": "The deployment's tenant id, normalized. Bound to the server."
        },
        "session_id": {
          "type": "string",
          "description": "The coding-agent session this belongs to (ADR 0002)."
        },
        "user_id": {
          "anyOf": [
            {
              "type": "string"
            },
            {
              "type": "null"
            }
          ],
          "description": "The developer identity when known. Null when capture omitted it."
        },
        "agent_harness": {
          "enum": [
            "claude-code",
            "copilot",
            "codex",
            "cursor",
            "pi"
          ],
          "type": "string",
          "description": "Which coding agent harness produced this fact."
        },
        "file_path": {
          "type": "string",
          "description": "The file edited, or empty string when the client emits none — common on rejects."
        },
        "accepted": {
          "type": "boolean",
          "description": "Whether the developer took the change."
        },
        "explicit": {
          "type": "boolean",
          "description": "True for a real human gesture; false for auto-applied or inferred. Implicit decisions move confidence but never classify a row."
        },
        "interaction_mode": {
          "enum": [
            "agent",
            "inline"
          ],
          "type": "string",
          "description": "Whether the edit interaction was `agent` or `inline`."
        },
        "commit_sha": {
          "anyOf": [
            {
              "type": "string"
            },
            {
              "type": "null"
            }
          ],
          "description": "The commit the client attributed, when it knew one."
        },
        "call_id": {
          "anyOf": [
            {
              "type": "string"
            },
            {
              "type": "null"
            }
          ],
          "description": "The provider's per-call id, joining this record back to the inference call that produced it. Null when the provider emits none."
        },
        "edit_retention_score": {
          "anyOf": [
            {
              "type": "number"
            },
            {
              "type": "null"
            }
          ],
          "minimum": 0.0,
          "maximum": 1.0,
          "description": "Graded preference strength in [0, 1]: how much applied edit text remains in a later file observation. Null unless capture or derivation supplies it."
        },
        "observation_delay_ms": {
          "anyOf": [
            {
              "type": "integer"
            },
            {
              "type": "null"
            }
          ],
          "minimum": 0,
          "description": "The delay before the edit retention score was observed. Null unless the harness reports one."
        },
        "occurred_at": {
          "type": "string",
          "format": "date-time",
          "description": "When the developer acted, stamped client-side. Required — it is what makes a redelivery collapse, and is never backfilled."
        },
        "captured_at": {
          "type": "string",
          "format": "date-time",
          "description": "When Sediment stored the fact, stamped server-side. Windows and ordering anchor on this, never on a client clock."
        },
        "raw": {
          "type": "object",
          "additionalProperties": {},
          "description": "The source payload after Basic redaction, kept so a translator defect stays recoverable. Shape varies by provider; never join on it."
        }
      },
      "additionalProperties": false,
      "required": [
        "decision_id",
        "org_id",
        "session_id",
        "user_id",
        "agent_harness",
        "file_path",
        "accepted",
        "explicit",
        "interaction_mode",
        "commit_sha",
        "call_id",
        "edit_retention_score",
        "observation_delay_ms",
        "occurred_at",
        "captured_at",
        "raw"
      ]
    },
    "CIOutcome": {
      "type": "object",
      "properties": {
        "schema_version": {
          "const": 1,
          "type": "integer",
          "description": "The normalized fact-contract version. CI outcomes use 1."
        },
        "outcome_id": {
          "type": "string",
          "description": "Stable id for this outcome."
        },
        "org_id": {
          "type": "string",
          "description": "The deployment's tenant id, normalized. Bound to the server."
        },
        "provider": {
          "enum": [
            "github_actions",
            "jenkins",
            "gitlab_ci",
            "circleci",
            "buildkite",
            "other"
          ],
          "type": "string",
          "description": "The normalized CI system selected by a capture adapter or declared by the bearer-authenticated vendor integration."
        },
        "run_id": {
          "type": "string",
          "description": "The provider-issued pipeline-run id, unique within the deployment organization and provider namespace."
        },
        "run_attempt": {
          "anyOf": [
            {
              "type": "integer"
            },
            {
              "type": "null"
            }
          ],
          "minimum": 1,
          "maximum": 9223372036854775807,
          "description": "The positive provider attempt number. Null when the provider omits it."
        },
        "repo": {
          "type": "string",
          "description": "`owner/repo`, lowercased. Empty string when the sender omitted it."
        },
        "commit_sha": {
          "type": "string",
          "description": "Full-length commit sha, lowercased."
        },
        "branch": {
          "type": "string",
          "description": "The branch the run or commit belongs to."
        },
        "result": {
          "enum": [
            "passed",
            "failed",
            "error",
            "timed_out",
            "cancelled",
            "skipped",
            "neutral",
            "unknown"
          ],
          "type": "string",
          "description": "Sediment's normalized terminal result."
        },
        "workflow_name": {
          "type": "string",
          "description": "The CI workflow's display name."
        },
        "workflow_id": {
          "anyOf": [
            {
              "type": "string"
            },
            {
              "type": "null"
            }
          ],
          "description": "The provider's stable workflow or pipeline-definition identity."
        },
        "workflow_path": {
          "anyOf": [
            {
              "type": "string"
            },
            {
              "type": "null"
            }
          ],
          "description": "The workflow definition's repo-relative path — the stable identity a display name does not give you. Null when unreported."
        },
        "run_url": {
          "anyOf": [
            {
              "type": "string"
            },
            {
              "type": "null"
            }
          ],
          "description": "Link to the run in the CI system. Null when unreported; never run identity."
        },
        "provider_result": {
          "anyOf": [
            {
              "type": "string"
            },
            {
              "type": "null"
            }
          ],
          "description": "The exact terminal value reported by the provider before normalization."
        },
        "error_type": {
          "anyOf": [
            {
              "type": "string"
            },
            {
              "type": "null"
            }
          ],
          "description": "A structured, low-cardinality provider error type. Null when absent."
        },
        "reason": {
          "anyOf": [
            {
              "type": "string"
            },
            {
              "type": "null"
            }
          ],
          "maxLength": 4096,
          "description": "A bounded, Basic-redacted provider reason. Null when absent."
        },
        "source_event_type": {
          "anyOf": [
            {
              "type": "string"
            },
            {
              "type": "null"
            }
          ],
          "description": "The source contract and event type. Null when unreported."
        },
        "source_spec_version": {
          "anyOf": [
            {
              "type": "string"
            },
            {
              "type": "null"
            }
          ],
          "description": "The source contract version. Null when unreported."
        },
        "source_event_id": {
          "anyOf": [
            {
              "type": "string"
            },
            {
              "type": "null"
            }
          ],
          "description": "The provider or standard source event identity. Null when unreported."
        },
        "pr_number": {
          "anyOf": [
            {
              "type": "integer"
            },
            {
              "type": "null"
            }
          ],
          "minimum": 1,
          "maximum": 9223372036854775807,
          "description": "The pull request the run belongs to, when there is one."
        },
        "captured_at": {
          "type": "string",
          "format": "date-time",
          "description": "When Sediment stored the fact, stamped server-side. Windows and ordering anchor on this, never on a client clock."
        },
        "raw": {
          "type": "object",
          "additionalProperties": {},
          "description": "The source payload after Basic redaction, kept so a translator defect stays recoverable. Shape varies by provider; never join on it."
        }
      },
      "additionalProperties": false,
      "required": [
        "schema_version",
        "outcome_id",
        "org_id",
        "provider",
        "run_id",
        "run_attempt",
        "repo",
        "commit_sha",
        "branch",
        "result",
        "workflow_name",
        "workflow_id",
        "workflow_path",
        "run_url",
        "provider_result",
        "error_type",
        "reason",
        "source_event_type",
        "source_spec_version",
        "source_event_id",
        "pr_number",
        "captured_at",
        "raw"
      ]
    },
    "Provenance": {
      "type": "object",
      "properties": {
        "policy_version": {
          "type": "string",
          "description": "The derivation implementation policy version."
        },
        "quarantine_revision": {
          "type": "integer",
          "description": "The integer quarantine-log high-water mark used for the derivation."
        },
        "policy_digest": {
          "anyOf": [
            {
              "type": "string"
            },
            {
              "type": "null"
            }
          ],
          "description": "SHA-256 of the fully resolved derivation policy when available."
        }
      },
      "additionalProperties": false,
      "required": [
        "policy_version",
        "quarantine_revision",
        "policy_digest"
      ]
    },
    "SessionAbandonment": {
      "type": "object",
      "properties": {
        "org_id": {
          "type": "string",
          "description": "The deployment's tenant id, normalized. Bound to the server."
        },
        "session_id": {
          "type": "string",
          "description": "The coding-agent session this belongs to (ADR 0002)."
        },
        "accepted_decisions": {
          "type": "integer",
          "description": "How many accepted edits this session had that never reached a commit."
        },
        "explicit_accepted_decisions": {
          "type": "integer",
          "description": "How many of those accepts came from a real human gesture. Only a session with at least one explicit accept can emit a negative attributed completion."
        },
        "last_decision_at": {
          "type": "string",
          "format": "date-time",
          "description": "The session's newest decision time — the anchor the grace horizon is measured back from."
        },
        "as_of": {
          "type": "string",
          "format": "date-time",
          "description": "The newest timestamp anywhere in the facts this derivation read. The horizon is measured against this, never the wall clock, so the same facts always yield the same verdict."
        },
        "provenance": {
          "$ref": "#/$defs/Provenance",
          "description": "The structured provenance for this artifact."
        }
      },
      "additionalProperties": false,
      "required": [
        "org_id",
        "session_id",
        "accepted_decisions",
        "explicit_accepted_decisions",
        "last_decision_at",
        "as_of",
        "provenance"
      ]
    }
  }
}