# Source register Checked on 19 September 2026. These are starting sources for the two research tasks, not an exhaustive bibliography. Mutable documentation should be version-pinned again when experiments are run. Source content has not been bundled or redistributed. ## S01 — Ask OKF architecture decision **Type:** Pinned project architecture **Source:** https://github.com/chris-page-gov/okf-explorer/blob/905e680f6d3ad385de9b8effc351566eba0ab2b3/docs/adr-ask-okf-context-assembly.md Scope and boundaries of deterministic context assembly. ## S02 — Ask OKF context types **Type:** Pinned implementation **Source:** https://github.com/chris-page-gov/okf-explorer/blob/905e680f6d3ad385de9b8effc351566eba0ab2b3/apps/okf-explorer/src/lib/context/types.ts The existing package, record, assertion and requirement types. ## S03 — Ask OKF context engine **Type:** Pinned implementation **Source:** https://github.com/chris-page-gov/okf-explorer/blob/905e680f6d3ad385de9b8effc351566eba0ab2b3/apps/okf-explorer/src/lib/context/index.ts Inspected definitions include budgets, validation, canonicalisation and phrase resolution; not a completed code audit. ## S04 — Ask OKF interface and page-tool verification **Type:** Project test receipt **Source:** https://github.com/chris-page-gov/okf-explorer/blob/905e680f6d3ad385de9b8effc351566eba0ab2b3/docs/ask-okf-ui-verification.md Reports local verification; explicitly separates deployment and actual AI-host invocation. ## S05 — Ask OKF pre-implementation gap analysis **Type:** Historical project analysis **Source:** https://github.com/chris-page-gov/okf-explorer/blob/905e680f6d3ad385de9b8effc351566eba0ab2b3/docs/ask-okf-gap-analysis-2026-09-16.md The recorded producer-modelling gap must not be mistaken for a present-day benefits ruling. ## S06 — Mike Stonebraker interview — publisher transcript **Type:** Primary interview transcript **Source:** https://www.developing.dev/p/turing-award-winner-postgres-disagreeing Published 20 April 2026. Interview opinions require evaluation, not automatic adoption. ## S07 — DBOS academic project **Type:** Research programme **Source:** https://dbos-project.github.io/ Academic publications and code, distinct from the product ecosystem. ## S08 — DBOS architecture **Type:** Current supplier documentation **Source:** https://docs.dbos.dev/architecture Workflow architecture, recovery, versioning and storage considerations. ## S09 — DBOS steps **Type:** Current supplier documentation **Source:** https://docs.dbos.dev/python/tutorials/step-tutorial Python step behaviour. Verify TypeScript equivalents before implementation. ## S10 — DBOS datasource transactions **Type:** Current supplier documentation **Source:** https://docs.dbos.dev/python/tutorials/transaction-tutorial Atomic recording of supported database transactions; not arbitrary external effects. ## S11 — Benchmarking How Workflow Execution Scales on Postgres **Type:** Supplier benchmark **Source:** https://www.dbos.dev/blog/benchmarking-workflow-execution-scalability-on-postgres No-op workflow and queue experiments, not BEP assembly or service-capacity measurements. ## S12 — Consistency and Correctness in Data-Oriented Workflow Systems **Type:** CIDR 2026 research paper **Source:** https://www.vldb.org/cidrdb/papers/2026/p9-stonebraker.pdf Workflow guarantees and a prototype. Not a blanket statement of released product functionality. ## S13 — OLTP Through the Looking Glass 16 Years Later: Communication is the New Bottleneck **Type:** CIDR 2025 research paper **Source:** https://vldb.org/cidrdb/papers/2025/p17-zhou.pdf End-to-end communication and isolation overheads in studied transaction workloads. ## S14 — Nanopublications **Type:** Project specification overview **Source:** https://nanopub.net/ Atomic assertions with provenance and publication metadata. ## S15 — PROV-DM: The PROV Data Model **Type:** W3C Recommendation **Source:** https://www.w3.org/TR/prov-dm/ Interoperable provenance modelling. ## S16 — Trusty URIs: Verifiable, Immutable, and Permanent Digital Artifacts for Linked Data **Type:** Research paper **Source:** https://arxiv.org/abs/1401.5775 Hash-bound linked artefacts; integrity and identity are not truth. ## S17 — RO-Crate 1.2 **Type:** Community specification **Source:** https://www.researchobject.org/ro-crate/specification/1.2/ Research-object packaging and linked metadata. ## S18 — ProvSQL **Type:** Primary project documentation **Source:** https://provsql.org/ Database query lineage; not automatic semantic validation of source assertions. ## S19 — Sufficient Context: A New Lens on Retrieval Augmented Generation Systems **Type:** Research paper; version 3, April 2025 **Source:** https://arxiv.org/abs/2411.06037 Separates insufficient evidence from failures to use available context. ## S20 — Measuring Attribution in Natural Language Generation Models **Type:** Research paper **Source:** https://arxiv.org/abs/2112.12870 Attributable to Identified Sources (AIS), including human-evaluation guidance. ## S21 — Enabling Large Language Models to Generate Text with Citations **Type:** Research paper **Source:** https://arxiv.org/abs/2305.14627 ALCE citation and generation evaluation. ## S22 — ChatGPT Voice **Type:** Current OpenAI product documentation **Source:** https://help.openai.com/en/articles/20001274 Distinguishes voice experiences and their capabilities; no Ask OKF invocation was tested. ## S23 — Connected apps in ChatGPT **Type:** Current OpenAI product documentation **Source:** https://help.openai.com/en/articles/11487775-connectors-in-chatgpt Voice support depends on the app and available features. ## S24 — GraphRAG local search **Type:** Primary project documentation **Source:** https://microsoft.github.io/graphrag/query/local_search/ Combines graph and text context; not equivalent to the proposed BEP guarantees. ## S25 — Lost in the Middle: How Language Models Use Long Contexts **Type:** Research paper **Source:** https://arxiv.org/abs/2307.03172 A reason to test evidence order and length on the actual consumer model. ## S26 — RECOMP: Improving Retrieval-Augmented LMs with Compression and Selective Augmentation **Type:** Research paper **Source:** https://arxiv.org/abs/2310.04408 Context compression and selective use as experimental comparators. ## S27 — From OKE Alerts to RCA Drafts: Building a Guardrailed OpenClaw Incident Assistant on OCI **Type:** Supplier reference implementation **Source:** https://blogs.oracle.com/cloud-infrastructure/oke-incident-postmortem-autopilot-openclaw-genai An explicit use of bounded evidence packages; not an interoperability standard. ## S28 — RFC 8785: JSON Canonicalization Scheme **Type:** IETF-stream informational RFC **Source:** https://www.rfc-editor.org/rfc/rfc8785 Canonical JSON representation; application-level array semantics still need definition. ## S29 — MCP Tools specification, 2026-07-28 **Type:** Protocol specification **Source:** https://modelcontextprotocol.io/specification/2026-07-28/server/tools Structured results and output schemas; client support remains a separate question. ## S30 — PostgreSQL: Sorting Rows **Type:** PostgreSQL official documentation **Source:** https://www.postgresql.org/docs/current/queries-order.html Explicit ordering is required where result order matters. ## S31 — PostgreSQL: Transaction Isolation **Type:** PostgreSQL official documentation **Source:** https://www.postgresql.org/docs/current/transaction-iso.html Read consistency and snapshot semantics. ## S32 — PostgreSQL: Materialized Views **Type:** PostgreSQL official documentation **Source:** https://www.postgresql.org/docs/current/rules-materializedviews.html Stored query results and refresh behaviour. ## S33 — Appropriate Uses For SQLite **Type:** SQLite official documentation **Source:** https://www.sqlite.org/whentouse.html Embedded storage and deployment trade-offs. ## S34 — DBOS TypeScript library **Type:** Product source repository **Source:** https://github.com/dbos-inc/dbos-transact-ts Repository metadata inspected; no library benchmark or comprehensive code audit performed.