MESA E2E Sertifikasyonu · v0.1.0
Data - Core Sözleşmesi
Bu rehberin Türkçe ve İngilizce sürümleri aynı konuyu açıklar. Bu sayfayla ilgili kaynak bölümleri en altta yer alır.
Bu belgenin kapsamı#
Profile B, veri ve ayarları sabitlenmiş Data → Core akışını bağımsız işlem alındılarıyla değerlendirir; doğrulanamayan koşullarda ilerlemez. Resmî çalıştırma ve ölçüm bileşenleri hazır, ancak B4/B6/B7/B8 için dış kanıtlar ve yeni bir resmî PASS sonucu henüz yoktur. Eski sonuçlar güncel sertifikasyon sayılmaz.
Aşağıda özgün belgeyi, komutları ve varsa test kayıtlarını kendi dilinde okuyabilirsiniz. Sonuçlar yalnızca belgede belirtilen tarih ve koşullar için geçerlidir.
Kaynak belgedeki ilgili bölümler
Bu bölüm, kaynak belgenin bu sayfayla ilgili kısımlarını değiştirmeden gösterir. Kaynak içindeki bir bölüme giden bağlantıyı açtığınızda metin otomatik olarak görünür.
Profile B Autonomous Agent Pack v3.1
Scope: MESA + MESA_Data native legal E2E certification on a dedicated certification branch.
Authority rule: the live checked-out repositories and runtime behavior outrank this document when code has changed. Never guess an API, CLI, schema, path, or capability: discover and record it first.
This is the most important cross-repository gate discovered during pack v2 design.
Definition of native#
A native PASS means the MESA_Data product publisher/engine, after normal MESA_Data release verification and authorization, causes the current MESA product to ingest the exact planned chunks through a supported MESA V4 product contract and truthfully tracks terminal mutation state.
The following do not satisfy native PASS:
- a standalone certification script calling
MesaV4Client.insert()for every chunk while bypassing MESA_Data's publisher; - writing MESA storage directly;
- importing MESA_Data's local staging DB;
- monkey-patching HTTP responses;
- changing qrels to whatever the bridge created.
A bridge may be used only to isolate downstream behavior and can lead at most to PROFILE_B_DIAGNOSTIC_BRIDGE_ONLY.
Required semantic lifecycle#
The exact implementation may change, but the native product contract must semantically achieve:
verified immutable MESA_Data release
→ deterministic SourceChunk list
→ authenticated MESA principal
→ authorized tenant/workspace/dataset/agent scope
→ valid active V4 session (if required by current MESA)
→ stable document/revision/chunk identity mapping
→ MESA cognitive ingestion of chunk content + source_ref
→ stable idempotency key supplied in the form MESA actually consumes
→ async mutation id captured
→ poll exact mutation endpoint/state machine
→ only terminal COMMITTED/legitimate duplicate counts as success
→ delivery ledger persists remote identity/state
→ later MESA search provenance maps back to MESA_Data source/version/chunkArchive mismatch to recheck#
In the supplied snapshots, MESA_Data mocked a generic /v4/sources/chunks route with a payload that did not include the session/revision fields required by MESA's /v4/memory/insert. MESA_Data sent Idempotency-Key as a header while the supplied MESA cognitive insert accepted idempotency_key in the body. The mocked /v4/health also did not match MESA's /health.
This is a likely P0 blocker until the live repos prove otherwise.
Repair strategy if still mismatched#
Do not force a predetermined code location. Inspect both architectures and choose the smallest correct product contract. Preferred direction, if compatible with current design, is to make MESA_Data's publisher understand the real MESA V4 lifecycle rather than inventing undocumented MESA endpoints.
The repair must address all required semantics, including:
- target health/capability discovery without guessing routes;
- principal credentials and RBAC bootstrap assumptions;
- session creation/retention if required;
- document/revision/chunk/source reference mapping;
- finalization ordering for multi-chunk revisions;
- idempotency placement and payload hash stability;
- mutation state normalization;
- response-loss/retry reconciliation;
- exact provenance metadata preserving original MESA_Data IDs and content hashes.
Do not weaken MESA authentication or create a public unauthenticated bulk-ingest endpoint merely to make the publisher easier.
Identity mapping rules#
Preserve MESA_Data identity. If MESA accepts the exact document_id, version_id as revision_id, and chunk_id, use them. If current identifier validators reject any form, create a deterministic collision-resistant mapping and include original IDs in metadata/source_ref. Never use random UUIDs for source identity.
A mapping table must be emitted in run evidence:
mesa_data_document_id → mesa_document_id
mesa_data_version_id → mesa_revision_id
mesa_data_chunk_id → mesa_chunk_id
content_hash → MESA provenance/content identityLive compatibility regression#
Mocks alone are insufficient. Before full corpus:
- start real local MESA production-equivalent service;
- create a tiny isolated MESA_Data release/canary through normal release code;
- run the native publisher engine;
- observe real HTTP status and mutation ID;
- wait for COMMITTED;
- search through public V4 API;
- assert returned provenance contains the canary's exact source identity;
- retry the same native delivery and assert no duplicate logical content/mutation side effects beyond legitimate idempotent/dedup state.
A focused mocked/unit regression may also be added to MESA_Data, but cannot replace this live gate.
Full delivery gate#
After H1, publish the frozen legal release using the exact same native code path as the canary. Compare:
planned chunks
new chunks to send
already committed/skipped chunks
remote COMMITTED chunks
failed/rejected chunks
MESA-visible source chunk/provenance countUnexplained mismatch = FAIL.
