公开记录项
JustAnIota.com is the IOTA-1 authority surface and current public WordPress workbench. Protocol5 is the .NET experiment track for approximate public-symbol retrieval.
公开记录项
The WordPress track publishes pages, demo registry data, REST conversion endpoints, validation surfaces, and browser review tools. The Protocol5 track now carries facade-led .NET/C#/SQL Server experiment work for phrase-first segmentation, public-symbol descriptor search, category-corpus lookup, seed-vs-SQL ranking demos, read-only API/status surfaces, embedding experiments, database-only fallback, and evaluation reports.
公开记录项
The two tracks must not blur claims: JustAnIota owns IOTA-1 authority language and public records, while Protocol5 must report approximate ranked results with provenance, scores, segment evidence, drift, and live-AI disclosure before any support language widens.
WordPress track
- Owns the canonical IOTA-1 public pages, discovery files, launch-readiness records, and reader-facing explanations.
- Runs the JustAnIota IOTA-1 Bidirectional Semantic Converter plugin and REST endpoints with a deterministic demo registry.
- Keeps browser fallbacks for review tools so local pages remain useful even when the REST service is unavailable.
- Receives Protocol5 evidence as reviewed vocabulary and documentation, not as a live SQL mutation dependency.
- Publishes validation reports and warnings instead of claiming certification or UAIX.org approval.
Protocol5 track
- Implements the .NET experiment with explicit facade contracts, logic services, ADO.NET repositories, SQL Server category corpus lookup, and optional LM Studio embedding assistance.
- Uses UCD, CLDR, Unihan, public metadata, approved grapheme or emoji sequences, Category.Categories, Category.Words, Category.ISO10646, and 英语 anchors as evidence sources.
- Exposes read-only public endpoints for status, conversion, meaning, similarity, round-trip, ranking demo, and category corpus search.
- Preserves database-only mode as a required fallback and discloses embedding or live-AI assistance when used.
- Returns ranked candidates with provenance, ranking lanes, evidence summaries, vector coverage, unknown rate, and drift metrics rather than promising exact translation.
- Keeps embedding population in local desktop or script tooling, not public web mutation routes.
Protocol5 API boundary
- Public status may report SQL corpus reachability, table rows, embedded rows, vector dimensions, seed count, and whether live embedding is configured.
- Public search and ranking-demo routes are evidence surfaces: they compare and query stored vectors but do not train, populate, approve, certify, or rewrite registries.
- Local population tooling owns embedding generation, model selection, dimensions, source-field choice, private connection strings, and write access to SQL vector columns.
- JustAnIota authority pages decide which durable evidence vocabulary becomes part of the IOTA-1 public record.
Technology advances to show without overclaiming
- Phrase-first segmentation and maximal stored-segment lookup as an explainable bridge between plain 英语 and compact candidates.
- Candidate ranking lanes that separate seed registry matches, category rows, word rows, ISO10646 public-symbol rows, and fallback evidence.
- Source-evidence atlas counts so a reviewer can see whether a candidate came from public Unicode assignment, plain-language descriptors, category corpus, word corpus, or 英语 anchors.
- Segment-vector summaries such as dimensions, magnitude, preview coordinates, vector coverage, and unknown rate when stored vectors are part of the experiment.
- Seed-only versus SQL-backed ranking demos that reveal when a richer corpus changes output or top-candidate choice.
- Database-only fallback and local-only population boundaries so public pages do not imply live training, public mutation, or exact translation.
Shared release rule
- Do not move roadmap ideas into public support language until code, docs, examples, tests, release notes, and discovery records agree.
- Do not use private-use characters, raw Unicode numbers, or model token IDs as semantic authority.
- Every public result should preserve mode, score, provenance, Unicode sequence, registry reference, segment trace, source atlas, vector evidence, warnings, and live-AI status when available.