eduevidence 6.0.0 → 6.3.0

package/schemas/v2/study.schema.json

@@ -10,7 +10,11 @@
     "independence_key", "identity_status", "extensions"
   ],
   "properties": {
-    "study_id": { "type": "string", "pattern": "^(STU-|STUDY-)" },
+    "study_id": {
+      "type": "string",
+      "pattern": "^(ST-|STU-|STUDY-)",
+      "description": "Study identity. ST- is the placeholder prefix some legacy/curated packs used before STU-/STUDY-; migration preserves original ids rather than renaming them."
+    },
     "source_ids": {
       "type": "array",
       "items": { "type": "string", "pattern": "^(SRC-|S-)" },

package/schemas/vNext/autoevolve-session.schema.json

@@ -1 +1,34 @@
-{"$schema":"http://json-schema.org/draft-07/schema#","type":"object","required":["session_id","status","max_experiments","promotion","experiments"],"properties":{"session_id":{"type":"string"},"status":{"enum":["initialized","running","plateau","budget_exhausted","completed","blocked"]},"max_experiments":{"type":"integer","minimum":1,"maximum":50},"max_cost_usd":{"type":"number","minimum":0},"max_wall_minutes":{"type":"integer","minimum":1},"promotion":{"const":"branch_only"},"experiments":{"type":"array","items":{"type":"string"}},"best_experiment_id":{"type":["string","null"]},"protected_integrity":{"type":"boolean"}},"additionalProperties":false}
+{
+  "$schema": "http://json-schema.org/draft-07/schema#",
+  "$id": "https://github.com/37chengshan/eduevidence/schemas/vNext/autoevolve-session.schema.json",
+  "title": "AutoevolveSessionReport",
+  "description": "Daily self-evolution session report (engine/autoevolve/runner.py -> state_root/daily-report.json). Records one bounded branch-only session: how many experiments ran, their promotion statuses, the stopping reason, cost and wall time, and whether runner-owned isolation was verified. Automatic KEEP requires verified OS isolation; evaluator self-attestation is ignored.",
+  "type": "object",
+  "additionalProperties": false,
+  "required": [
+    "run_tag", "branch", "experiments", "statuses", "cost", "wall_minutes",
+    "plateau", "stop_reason", "promotion", "holdout_isolation_verified", "mutation_view"
+  ],
+  "properties": {
+    "run_tag": {"type": "string", "minLength": 1, "description": "Session tag; also the SkillExperiment.session_id of every experiment in this session."},
+    "branch": {"type": "string", "minLength": 1, "description": "Branch the session ran on. Promotion is branch-only."},
+    "experiments": {"type": "integer", "minimum": 0, "description": "Number of experiments attempted."},
+    "statuses": {"type": "array", "items": {"enum": ["created", "KEEP", "REJECT", "RETEST", "HUMAN_REVIEW", "CRASH", "INVALID"]}, "description": "Per-experiment promotion verdict, in attempt order."},
+    "best_experiment_id": {"type": ["string", "null"]},
+    "best_candidate_commit": {"type": ["string", "null"]},
+    "cost": {"type": "number", "minimum": 0, "description": "Cumulative session cost in USD, as reported by the agent and evaluator."},
+    "wall_minutes": {"type": "number", "minimum": 0},
+    "plateau": {"type": "boolean"},
+    "stop_reason": {"type": "string", "minLength": 1, "description": "Why the loop stopped (for example budget_exhausted, plateau, completed)."},
+    "promotion": {"const": "branch_only", "description": "Daily mode never promotes directly; KEEP only marks a candidate branch."},
+    "branch_push_requested": {"type": "boolean"},
+    "branch_pushed": {"type": "boolean"},
+    "mutation_view": {"type": "string", "minLength": 1, "description": "Isolation actually used for the mutation view."},
+    "holdout_isolation_verified": {"type": "boolean", "description": "Runner-owned OS isolation verification. Automatic KEEP requires true."},
+    "isolation_provider": {"type": "string"},
+    "isolation_reason": {"type": "string"},
+    "eval_suite_hash": {"type": "string", "description": "Trusted evaluation suite hash captured before the session; a mid-session change invalidates the experiment."},
+    "security_note": {"type": "string"},
+    "candidate_artifacts": {"type": "string", "description": "Where candidate artifacts live; local session state is never auto-pushed."}
+  }
+}

package/schemas/vNext/eval-snapshot.schema.json

@@ -1 +1,77 @@
-{"$schema":"http://json-schema.org/draft-07/schema#","type":"object","required":["eval_id","hard_gates_passed","science_score","research_score","robustness","cost","latency","complexity","repeats","noise_floor","dev_passed","holdout_passed","adversarial_passed","holdout_isolation_verified","eval_suite_hash"],"properties":{"eval_id":{"type":"string"},"hard_gates_passed":{"type":"boolean"},"science_score":{"type":"number"},"research_score":{"type":"number"},"robustness":{"type":"number"},"cost":{"type":"number","minimum":0},"latency":{"type":"number","minimum":0},"complexity":{"type":"number","minimum":0},"repeats":{"type":"integer","minimum":1},"noise_floor":{"type":"number","minimum":0},"dev_passed":{"type":"boolean"},"holdout_passed":{"type":"boolean"},"adversarial_passed":{"type":"boolean"},"holdout_isolation_verified":{"type":"boolean"},"eval_suite_hash":{"type":"string","minLength":1}},"additionalProperties":false}
+{
+  "$schema": "http://json-schema.org/draft-07/schema#",
+  "title": "EvalSnapshot",
+  "description": "The measurement record for one self-evolution candidate (engine/autoevolve/core.py). Carries gate outcomes, science/research scores, robustness, cost, latency, repeats and the trusted eval-suite hash. Runner-owned snapshots ignore evaluator self-attestation; a mid-session suite-hash change invalidates the experiment.",
+  "type": "object",
+  "required": [
+    "eval_id",
+    "hard_gates_passed",
+    "science_score",
+    "research_score",
+    "robustness",
+    "cost",
+    "latency",
+    "complexity",
+    "repeats",
+    "noise_floor",
+    "dev_passed",
+    "holdout_passed",
+    "adversarial_passed",
+    "holdout_isolation_verified",
+    "eval_suite_hash"
+  ],
+  "properties": {
+    "eval_id": {
+      "type": "string"
+    },
+    "hard_gates_passed": {
+      "type": "boolean"
+    },
+    "science_score": {
+      "type": "number"
+    },
+    "research_score": {
+      "type": "number"
+    },
+    "robustness": {
+      "type": "number"
+    },
+    "cost": {
+      "type": "number",
+      "minimum": 0
+    },
+    "latency": {
+      "type": "number",
+      "minimum": 0
+    },
+    "complexity": {
+      "type": "number",
+      "minimum": 0
+    },
+    "repeats": {
+      "type": "integer",
+      "minimum": 1
+    },
+    "noise_floor": {
+      "type": "number",
+      "minimum": 0
+    },
+    "dev_passed": {
+      "type": "boolean"
+    },
+    "holdout_passed": {
+      "type": "boolean"
+    },
+    "adversarial_passed": {
+      "type": "boolean"
+    },
+    "holdout_isolation_verified": {
+      "type": "boolean"
+    },
+    "eval_suite_hash": {
+      "type": "string",
+      "minLength": 1
+    }
+  },
+  "additionalProperties": false
+}

package/schemas/vNext/execution-plan.schema.json

@@ -1 +1,50 @@
-{"$schema":"http://json-schema.org/draft-07/schema#","type":"object","required":["complexity","tasks","max_parallel_workers","parallel_groups","plan_id"],"properties":{"complexity":{"enum":["S","M","L"]},"tasks":{"type":"array","items":{"$ref":"task-spec.schema.json"}},"max_parallel_workers":{"type":"integer","minimum":0,"maximum":6},"parallel_groups":{"type":"array","items":{"type":"array","minItems":1,"items":{"type":"string"}}},"plan_id":{"type":["string","null"]}},"additionalProperties":false}
+{
+  "$schema": "http://json-schema.org/draft-07/schema#",
+  "title": "ExecutionPlan",
+  "description": "The plan for one research run (engine/orchestration.py ExecutionPlanner.plan). Complexity level, the ordered TaskSpecs, the maximum parallel workers and the parallel groups; S delegates zero tasks, M and L are capped by policy. scripts/check_autoresearch_invariants.py enforces those caps.",
+  "type": "object",
+  "required": [
+    "complexity",
+    "tasks",
+    "max_parallel_workers",
+    "parallel_groups",
+    "plan_id"
+  ],
+  "properties": {
+    "complexity": {
+      "enum": [
+        "S",
+        "M",
+        "L"
+      ]
+    },
+    "tasks": {
+      "type": "array",
+      "items": {
+        "$ref": "task-spec.schema.json"
+      }
+    },
+    "max_parallel_workers": {
+      "type": "integer",
+      "minimum": 0,
+      "maximum": 6
+    },
+    "parallel_groups": {
+      "type": "array",
+      "items": {
+        "type": "array",
+        "minItems": 1,
+        "items": {
+          "type": "string"
+        }
+      }
+    },
+    "plan_id": {
+      "type": [
+        "string",
+        "null"
+      ]
+    }
+  },
+  "additionalProperties": false
+}

package/schemas/vNext/gap-priority.schema.json

@@ -1 +1,54 @@
-{"$schema":"http://json-schema.org/draft-07/schema#","type":"object","required":["gap_id","dvi_band","cost_band","decision_material","drivers","next_research_mode","score"],"properties":{"gap_id":{"type":"string"},"dvi_band":{"enum":["HIGH","MEDIUM","LOW"]},"cost_band":{"enum":["HIGH","MEDIUM","LOW"]},"decision_material":{"type":"boolean"},"drivers":{"type":"array","items":{"type":"string"}},"next_research_mode":{"enum":["secondary_evidence_search","defer","empirical_evidence_needed"]},"score":{"type":"integer"}},"additionalProperties":false}
+{
+  "$schema": "http://json-schema.org/draft-07/schema#",
+  "title": "GapPriority",
+  "description": "Ranking of one knowledge gap for research attention (engine/autoresearch/gap_priority.py). Combines decision sensitivity, current uncertainty, directness deficit, applicability, availability and risk minus cost into a score, and names the next research mode. Produced by controllers.select_gap(), not by an LLM.",
+  "type": "object",
+  "required": [
+    "gap_id",
+    "dvi_band",
+    "cost_band",
+    "decision_material",
+    "drivers",
+    "next_research_mode",
+    "score"
+  ],
+  "properties": {
+    "gap_id": {
+      "type": "string"
+    },
+    "dvi_band": {
+      "enum": [
+        "HIGH",
+        "MEDIUM",
+        "LOW"
+      ]
+    },
+    "cost_band": {
+      "enum": [
+        "HIGH",
+        "MEDIUM",
+        "LOW"
+      ]
+    },
+    "decision_material": {
+      "type": "boolean"
+    },
+    "drivers": {
+      "type": "array",
+      "items": {
+        "type": "string"
+      }
+    },
+    "next_research_mode": {
+      "enum": [
+        "secondary_evidence_search",
+        "defer",
+        "empirical_evidence_needed"
+      ]
+    },
+    "score": {
+      "type": "integer"
+    }
+  },
+  "additionalProperties": false
+}

package/schemas/vNext/negative-search-record.schema.json

@@ -1 +1,68 @@
-{"$schema":"http://json-schema.org/draft-07/schema#","type":"object","required":["negative_search_id","research_iteration_id","gap_id","queries","providers","candidate_count","fetched_count","eligible_count","conclusion"],"properties":{"negative_search_id":{"type":"string"},"research_iteration_id":{"type":"string"},"gap_id":{"type":"string"},"queries":{"type":"array","items":{"type":"string"}},"providers":{"type":"array","items":{"type":"string"}},"candidate_count":{"type":"integer","minimum":0},"fetched_count":{"type":"integer","minimum":0},"eligible_count":{"const":0},"exclusion_reasons":{"type":"object","additionalProperties":{"type":"integer","minimum":0}},"scope":{"type":"object"},"searched_at":{"type":"string"},"conclusion":{"const":"no_eligible_evidence_found_within_search_scope"}},"additionalProperties":false}
+{
+  "$schema": "http://json-schema.org/draft-07/schema#",
+  "title": "NegativeSearchRecord",
+  "description": "A search that legitimately returned nothing usable (engine/autoresearch/contracts.py NegativeSearchRecord). Zero-result searches are recorded rather than discarded: without them, absence of evidence is indistinguishable from failure to look, and retrieval saturation cannot be judged. Written by ResearchMemory.append_negative_search().",
+  "type": "object",
+  "required": [
+    "negative_search_id",
+    "research_iteration_id",
+    "gap_id",
+    "queries",
+    "providers",
+    "candidate_count",
+    "fetched_count",
+    "eligible_count",
+    "conclusion"
+  ],
+  "properties": {
+    "negative_search_id": {
+      "type": "string"
+    },
+    "research_iteration_id": {
+      "type": "string"
+    },
+    "gap_id": {
+      "type": "string"
+    },
+    "queries": {
+      "type": "array",
+      "items": {
+        "type": "string"
+      }
+    },
+    "providers": {
+      "type": "array",
+      "items": {
+        "type": "string"
+      }
+    },
+    "candidate_count": {
+      "type": "integer",
+      "minimum": 0
+    },
+    "fetched_count": {
+      "type": "integer",
+      "minimum": 0
+    },
+    "eligible_count": {
+      "const": 0
+    },
+    "exclusion_reasons": {
+      "type": "object",
+      "additionalProperties": {
+        "type": "integer",
+        "minimum": 0
+      }
+    },
+    "scope": {
+      "type": "object"
+    },
+    "searched_at": {
+      "type": "string"
+    },
+    "conclusion": {
+      "const": "no_eligible_evidence_found_within_search_scope"
+    }
+  },
+  "additionalProperties": false
+}

package/schemas/vNext/research-iteration.schema.json

@@ -1 +1,87 @@
-{"$schema":"http://json-schema.org/draft-07/schema#","type":"object","required":["iteration_id","project_id","base_graph_revision","gap_id","strategy","status"],"properties":{"iteration_id":{"type":"string"},"project_id":{"type":"string"},"base_graph_revision":{"type":"integer","minimum":0},"gap_id":{"type":"string"},"gap_lineage_key":{"type":["string","null"],"pattern":"^KGK-"},"strategy":{"type":"object"},"validated_evidence_ids":{"type":"array","items":{"type":"string"}},"negative_search_ids":{"type":"array","items":{"type":"string"}},"evidence_gain":{"type":"object"},"new_graph_revision":{"type":["integer","null"]},"decision_snapshot_id":{"type":["string","null"]},"status":{"enum":["completed_gain","completed_no_gain","search_saturated","empirical_needed","budget_exhausted","tool_failure","invalid"]},"started_at":{"type":"string"},"completed_at":{"type":["string","null"]}},"additionalProperties":true}
+{
+  "$schema": "http://json-schema.org/draft-07/schema#",
+  "title": "ResearchIteration",
+  "description": "One autoresearch loop iteration (engine/autoresearch/contracts.py ResearchIteration, persisted by ResearchMemory.append_iteration). Records which knowledge gap was attacked, the strategy and budget used, what was retrieved and validated, whether a new GraphRevision was created, and how the loop stopped. Written by scripts/research_auto_cli.py.",
+  "type": "object",
+  "required": [
+    "iteration_id",
+    "project_id",
+    "base_graph_revision",
+    "gap_id",
+    "strategy",
+    "status"
+  ],
+  "properties": {
+    "iteration_id": {
+      "type": "string"
+    },
+    "project_id": {
+      "type": "string"
+    },
+    "base_graph_revision": {
+      "type": "integer",
+      "minimum": 0
+    },
+    "gap_id": {
+      "type": "string"
+    },
+    "gap_lineage_key": {
+      "type": [
+        "string",
+        "null"
+      ],
+      "pattern": "^KGK-"
+    },
+    "strategy": {
+      "type": "object"
+    },
+    "validated_evidence_ids": {
+      "type": "array",
+      "items": {
+        "type": "string"
+      }
+    },
+    "negative_search_ids": {
+      "type": "array",
+      "items": {
+        "type": "string"
+      }
+    },
+    "evidence_gain": {
+      "type": "object"
+    },
+    "new_graph_revision": {
+      "type": [
+        "integer",
+        "null"
+      ]
+    },
+    "decision_snapshot_id": {
+      "type": [
+        "string",
+        "null"
+      ]
+    },
+    "status": {
+      "enum": [
+        "completed_gain",
+        "completed_no_gain",
+        "search_saturated",
+        "empirical_needed",
+        "budget_exhausted",
+        "tool_failure",
+        "invalid"
+      ]
+    },
+    "started_at": {
+      "type": "string"
+    },
+    "completed_at": {
+      "type": [
+        "string",
+        "null"
+      ]
+    }
+  },
+  "additionalProperties": true
+}

package/schemas/vNext/research-strategy.schema.json

@@ -1 +1,62 @@
-{"$schema":"http://json-schema.org/draft-07/schema#","type":"object","required":["strategy_id","experiment_type","hypothesis","expected_gain","budget"],"properties":{"strategy_id":{"type":"string","minLength":1},"experiment_type":{"enum":["TARGETED_RETRIEVAL","COUNTER_EVIDENCE_RETRIEVAL","APPLICABILITY_RETRIEVAL","TEMPORAL_REFRESH","CITATION_CHAINING","SCREENING_PRIORITY","SOURCE_RECOVERY"]},"hypothesis":{"type":"string","minLength":1},"expected_gain":{"type":"string","minLength":1},"budget":{"type":"object","required":["max_queries","max_candidates","max_fulltext_fetches"],"properties":{"max_queries":{"type":"integer","minimum":0},"max_candidates":{"type":"integer","minimum":0},"max_fulltext_fetches":{"type":"integer","minimum":0}},"additionalProperties":false}},"additionalProperties":false}
+{
+  "$schema": "http://json-schema.org/draft-07/schema#",
+  "title": "ResearchStrategy",
+  "description": "The planned attack on one knowledge gap (engine/autoresearch/contracts.py ResearchStrategy). Names the experiment type, the hypothesis it tests, the expected gain, and the research budget. strategy_id/hypothesis/expected_gain are required by ResearchStrategy.validate().",
+  "type": "object",
+  "required": [
+    "strategy_id",
+    "experiment_type",
+    "hypothesis",
+    "expected_gain",
+    "budget"
+  ],
+  "properties": {
+    "strategy_id": {
+      "type": "string",
+      "minLength": 1
+    },
+    "experiment_type": {
+      "enum": [
+        "TARGETED_RETRIEVAL",
+        "COUNTER_EVIDENCE_RETRIEVAL",
+        "APPLICABILITY_RETRIEVAL",
+        "TEMPORAL_REFRESH",
+        "CITATION_CHAINING",
+        "SCREENING_PRIORITY",
+        "SOURCE_RECOVERY"
+      ]
+    },
+    "hypothesis": {
+      "type": "string",
+      "minLength": 1
+    },
+    "expected_gain": {
+      "type": "string",
+      "minLength": 1
+    },
+    "budget": {
+      "type": "object",
+      "required": [
+        "max_queries",
+        "max_candidates",
+        "max_fulltext_fetches"
+      ],
+      "properties": {
+        "max_queries": {
+          "type": "integer",
+          "minimum": 0
+        },
+        "max_candidates": {
+          "type": "integer",
+          "minimum": 0
+        },
+        "max_fulltext_fetches": {
+          "type": "integer",
+          "minimum": 0
+        }
+      },
+      "additionalProperties": false
+    }
+  },
+  "additionalProperties": false
+}

package/schemas/vNext/skill-experiment.schema.json

@@ -1 +1,90 @@
-{"$schema":"http://json-schema.org/draft-07/schema#","type":"object","required":["experiment_id","session_id","parent_skill_revision","hypothesis","mutation_scope","status"],"properties":{"experiment_id":{"type":"string"},"session_id":{"type":"string"},"parent_skill_revision":{"type":"string"},"hypothesis":{"type":"string","minLength":1},"mutation_scope":{"type":"array","items":{"type":"string"},"minItems":1},"changed_files":{"type":"array","items":{"type":"string"}},"candidate_commit":{"type":["string","null"]},"baseline_eval_id":{"type":["string","null"]},"candidate_eval_id":{"type":["string","null"]},"protected_hash_before":{"type":["string","null"]},"protected_hash_after":{"type":["string","null"]},"status":{"enum":["created","KEEP","REJECT","RETEST","HUMAN_REVIEW","CRASH","INVALID"]},"promotion_reason":{"type":"string"},"complexity_delta":{"type":"number"}},"additionalProperties":false}
+{
+  "$schema": "http://json-schema.org/draft-07/schema#",
+  "title": "SkillExperiment",
+  "description": "One bounded self-evolution experiment (engine/autoevolve/core.py). Pairs a hypothesis and mutation scope with the baseline and candidate evaluation ids and the protected-tree hashes before and after, then records the promotion verdict. Automatic KEEP requires runner-owned isolation and an intact protected tree.",
+  "type": "object",
+  "required": [
+    "experiment_id",
+    "session_id",
+    "parent_skill_revision",
+    "hypothesis",
+    "mutation_scope",
+    "status"
+  ],
+  "properties": {
+    "experiment_id": {
+      "type": "string"
+    },
+    "session_id": {
+      "type": "string"
+    },
+    "parent_skill_revision": {
+      "type": "string"
+    },
+    "hypothesis": {
+      "type": "string",
+      "minLength": 1
+    },
+    "mutation_scope": {
+      "type": "array",
+      "items": {
+        "type": "string"
+      },
+      "minItems": 1
+    },
+    "changed_files": {
+      "type": "array",
+      "items": {
+        "type": "string"
+      }
+    },
+    "candidate_commit": {
+      "type": [
+        "string",
+        "null"
+      ]
+    },
+    "baseline_eval_id": {
+      "type": [
+        "string",
+        "null"
+      ]
+    },
+    "candidate_eval_id": {
+      "type": [
+        "string",
+        "null"
+      ]
+    },
+    "protected_hash_before": {
+      "type": [
+        "string",
+        "null"
+      ]
+    },
+    "protected_hash_after": {
+      "type": [
+        "string",
+        "null"
+      ]
+    },
+    "status": {
+      "enum": [
+        "created",
+        "KEEP",
+        "REJECT",
+        "RETEST",
+        "HUMAN_REVIEW",
+        "CRASH",
+        "INVALID"
+      ]
+    },
+    "promotion_reason": {
+      "type": "string"
+    },
+    "complexity_delta": {
+      "type": "number"
+    }
+  },
+  "additionalProperties": false
+}