v0.2.0: Complete term slices and publish the OEIS guide

Complete term slices and publish the OEIS guide

• Publish the OEIS guide at skill://oeis-guide/SKILL.md for candidate verification and partial-result interpretation.
• oeis_get_terms serves slices by count up to 1,000 terms (#21).
• oeis_get_terms can fill a slice from the next b-file part and preserves the cursor when an optional read cannot establish later terms (#23).
• oeis_get_sequence renders program language tags literally, including underscores and trailing hashes (#19).
• oeis_list_reference explains that quoted searches match normalized tokens and points callers to the entry formulas to verify signs and operators (#22).
• The desktop bundle retains the published guide and omits development files. The root plugin manifest supports Agent Plugins discovery, and Bun builds use 1.4.4.
• Dependencies — @cyanheads/mcp-ts-core ^0.13.15 → ^0.13.22, @biomejs/biome 2.5.15 → 2.6.6, @modelcontextprotocol/server ^2.3.1 → ^2.4.0, @types/node 26.6.5 → 26.6.6, @vitest/coverage-istanbul 5.0.4 → 5.1.0, ignore ^7.0.12 → ^7.0.14, marked ^18.1.1 → ^18.2.0, tsc-alias ^1.9.7 → ^1.9.10, vitest ^5.0.4 → ^5.1.0

CHANGELOG v0.2.0 · release PR #24

v0.1.3: Search split notice, b-file fixes, contributor text as code

Search split notice, b-file fixes, contributor text as code

• oeis_search_sequences notice when OEIS split numbers into separate searches, with the comma-joined run to send instead (#13)
• oeis_search_sequences zero-hit _ wildcard run gets the term-run advice (#18)
• oeis_get_terms drops a cached no-b-file answer once a fetched record lists a b-file (#12)
• oeis_get_terms skips and counts malformed b-file lines; a b-file with no n a(n) pair fails as upstream_unparseable and is not cached (#16)
• oeis_get_terms serves a limit above 1000 at 1000 with a notice (#17)
• content[] renders contributor text as code spans and fences (#15)
• Framework sync to mcp-ts-core 0.13.15 (skills, scripts, Dockerfile, CodeQL, server.json)
• deps: @cyanheads/mcp-ts-core ^0.13.10 → ^0.13.15 (+ dev-dep bumps)

CHANGELOG v0.1.3 · release PR #20

v0.1.2: Paged b-file reads, section paging, richer summary rows

Paged b-file reads, section paging, richer summary rows

• oeis_get_terms reads b-file terms past the first 1 MiB, in ranged 1 MiB parts, at most two part reads a call (#6)
• oeis_get_sequence pages a sections selection past the 100,000-byte response budget; fromItem resumes from nextFromItem (#7)
• Search, identify, and cross-refs rows carry author, modified, and legacyIds (#8)
• A fresh search row with a later edit expires the cached record, and a row listing an entry drops its remembered 404 (#11)
• Zero-hit term-match notices name the data-line limit (#9, #10); server.json lists the hosted endpoint; linear-time tests bound thread CPU time
• deps: @modelcontextprotocol/server ^2.1.0 → ^2.2.0 (the installed SDK, shared with @cyanheads/mcp-ts-core, is now 2.2.0)

CHANGELOG v0.1.2 · release PR #14

v0.1.1: OEIS integer sequence lookup, identification, and terms

OEIS integer sequence lookup, identification, and terms

• oeis_identify_sequence: identify sequences from a run of up to 60 consecutive terms, with the index where the run begins in each (#4)
• oeis_search_sequences: search in OEIS query syntax, sorted by relevance, number, created, or modified
• oeis_get_sequence: one entry with formulas, programs, and its other sections, outlined past 24,000 characters (#2)
• oeis_get_terms: b-file or data-line terms as exact decimal strings, sliced by count and bytes (#1, #3)
• oeis_get_cross_refs: the sequences an entry cross-references, with their notes, or the entries that mention it
• oeis_list_reference: keyword flags, search syntax, and identifier conventions from static tables
• oeis://sequence/{aNumber}: the whole entry as JSON
• oeis.org paced to one request every 10 seconds, with retries and caching; OEIS_QUEUE_MAX_WAIT_MS caps the queue wait
• OEIS content is CC BY-SA 4.0, and every record and row carries its oeis.org URL

CHANGELOG v0.1.1 · release PR #5