Linked Data REPL experiment

This is a small Codex Desktop experiment for running in-memory RDF/SPARQL work with a user-owned persistent JavaScript REPL. The local Communica engine and RDF store belong to a worker task; a conversational coordinator delegates a bounded task and receives its evidence-backed result.

Current Codex runtime boundary

The active project configuration registers the separately saved cleanroom_node_repl MCP from /Users/cvardema/dev/git/LA3D/linked-science-cloud/node-repl-network-probe. It is a user-owned CodeAct runtime with persistent JavaScript, RLM context operations, broker-owned PEEK orientation, and the observed three-tool contract. This project does not use or configure Codex Desktop's bundled node_repl.

The former restricted network profile remains disabled at .codex/config.restricted-profile.toml.disabled for historical reference. It is not active configuration and must not be re-enabled to grant the bundled REPL network access. Tool exposure, project-root selection, module resolution, persistence, RLM state, PEEK state, and guarded network reachability remain separate properties that must be observed rather than inferred.

Worker pattern

1. The coordinator keeps the conversation, scope, and approval boundary.
2. A fresh Local task loads the project-scoped cleanroom_node_repl, verifies its cwd, and bootstraps the Linked Science facade once per kernel.
3. The worker verifies a query in a second REPL call and reports only what was actually observed. Kernel reset discards bindings, RLM context, and resident handles while broker PEEK orientation remains advisory.

Run the disposable local check with:

npm run smoke

For the tested REPL initialization and persistence check, use the project-local Linked Data REPL skill.

Linked Science runtime

The production-oriented local adapter is bootstrapped by lib/cleanroom-linked-science-bootstrap.mjs and implemented by lib/linked-science-runtime.mjs. It installs one stable CodeAct-style JavaScript facade, validates this checkout and its declared dependency root, registers discovery material with clean-room RLM context, keeps Communica private as the query kernel, delegates compact orientation to broker-owned PEEK, and retains ontology/schema/SHACL/instance graphs and results behind epoch-bearing handles. Start with runtime discovery; the checked-in machine API schema supports fresh-agent lookup. The local slice adds no live endpoint or recursive model access.

Capability map and next work

• Synthetic session: local in-memory RDF is materialized under a symbolic handle and can be profiled, paged, or derived without printing the full result.
• CodeAct runtime: persistent model-written JavaScript uses the stable linkedScience facade for local graph objects, schema search, Communica queries, generic derivation, bounded views, PEEK orientation, and explicit stale-handle recovery.
• Guarded live table: the separate Identifiers.org demonstration profile pins one SPARQL endpoint and permits only bounded read-only registry tables through Communica.
• Atomic typed live retention: queryToHandleGuarded preflights the session and handle before network access, runs one guarded read operation, retains bindings, boolean, or quad results, and returns one compact receipt containing all transport attempts and the result shape.
• Open-goal operation choice: two fresh live UniProt workers used the same guarded surface without query templates. A factual goal produced bounded SELECT handles; a graph-shaped goal independently produced a six-quad CONSTRUCT handle. See the operation-selection experiment and its compact receipts.
• Guarded federation: the uniprotRheaWikidataFederation profile starts at UniProt and permits bounded SELECT federation only to the exact Rhea and Wikidata SPARQL endpoints. It remains read-only, result-capped, redirect-free, and profile-bound.
• WikiPathways and ChEBI: the wikiPathwaysRead profile permits bounded reads from the exact WikiPathways endpoint, including ChEBI-linked data nodes. Exact ChEBI compound enrichment uses a validated official EMBL-EBI JSON URL per accession; ChEBI is not modeled as a SPARQL service and its full ontology is not fetched by default.
• Evidence manifests: a small versioned manifest points a worker to authoritative ontology, examples, named-graph description, and an approved query-policy reference. It is not a planner or fallback query library. See the goal-loop dossier.
• Source orientation index: resources/index.md is a deliberately small directory of scientific roles, candidate entry points, and identifier anchors. It helps an agent choose where to investigate; it is not endpoint code, an allowlist, a health claim, or a query plan.
• Documentation discovery: a separate, pinned UniProt schema-document profile retrieves the canonical rendered RDF schema with compact hash provenance. It does not permit data-endpoint queries. See the schema-discovery experiment.
• Prior grounding evaluation: workers may use pretrained knowledge as hypotheses, but must ground, correct, reject, or leave it unresolved before it enters a plan. The evaluation separately diagnoses methodology, tool-surface, environment, and source failures. See the dossier.
• Adaptive evidence acquisition: an exact approved ontology or document can be content-sniffed and retained behind a handle even when its suffix and HTTP metadata disagree. Failed requests become typed attempt evidence, not claims of absence.
• Symbolic orientation cache: a bounded PEEK-aligned map retains stable sources, parsing facts, failures, identifiers, and reusable result handles while bulk evidence remains in the REPL. It is not a task-answer store or a competing workflow engine.
• Codex-goal-compatible evidence state: the earlier checklist-style worker guidance is transitional. Codex owns goals and worker lifecycle; the first project-local slice now attaches guarded acquisition and symbolic orientation events to retained Linked Data state. See the decision and experiment dossier.
• Presentation handoff: a retained handle can yield a typed, bounded table model with source handle and compact provenance. It is not HTML, a full result, or a UniProt record browser.

Use the agent context router to load project guidance by task type. Read the living roadmap, active task queue, dated work logs, experiment dossiers, and generated context-map receipts for scope and evidence. The clean-worker map evaluation now has one successful constrained local run: a real REPL preflight, retained-handle reuse under coordinator steering, and reset honesty. It does not establish open-ended or endpoint navigation. The large-result export protocol remains documentation only, with no exporter or generated export artifact.

Current live-navigation boundary

Live use remains explicitly approved and profile-bound. Named SPARQL profiles pin Identifiers.org, UniProt, WikiPathways, Rhea, and Wikidata as described above. Exact ChEBI compound acquisition pins one validated public EMBL-EBI API URL at a time. These surfaces permit bounded read-only operations and recorded transport provenance only. They do not authorize arbitrary provider URLs, other REST routes or endpoints, mutation-capable methods, SERVICE SILENT, or unbounded result handling. Do not substitute another endpoint or runtime without explicit approval.