fix(models): surface Kimi Coding k3 when searching kimi

[HexLab98]

Summary

• Kimi Coding discovers its flagship as wire id k3 (not kimi-k3). Model pickers (dashboard MoA / Change, TUI, CLI, desktop) fuzzy-filtered on that id only, so typing kimi hid k3 while every sibling kimi-* model stayed visible.
• Add picker-only search aliases (k3 ↔ kimi-k3 / kimi) without changing the wire id or putting k3 in the static catalog (that approach was deliberately rejected in feat(models): add k3 to kimi-coding and kimi-coding-cn model lists #65969 / feat(kimi): discover K3 on the Kimi Coding endpoint (salvage #65725) #65922).
• Cover the haystack + fuzzy filter with Python and TUI tests.

Test plan

• scripts/run_tests.sh tests/hermes_cli/test_model_search.py tests/hermes_cli/test_curses_ui_fuzzy_rank.py -q
• cd ui-tui && npm test -- --run src/lib/model-search-text.test.ts
• Dashboard → Models → MoA → Change → Kimi For Coding / Kimi Coding Plan → search kimi → confirm k3 appears
• Same picker → search k3 → still finds it
• With a Coding Plan key, clear search → k3 still only appears via live discovery (not a static catalog entry)

[alt-glitch]

This was generated by AI during triage.

Related to merged #65922 and closed #65969: this PR leaves K3 discovery and its wire ID unchanged, adding only synchronized picker-search aliases across the UI surfaces.

[GottZ]

This was generated by AI during triage.

Summary

Two PRs address the same picker-search failure: #67409 adds picker-only aliases so the live k3 wire ID matches “kimi” across CLI, TUI, web, and desktop, while #68153 carries that fix onto current main and additionally deduplicates live k3 against curated kimi-k3.

Related pull requests

• #67409 related — (+243/-8) — superseded duplicate: Adds synchronized k3 search aliases across all picker surfaces without changing the wire ID or adding a static catalog entry, directly addressing the reported search failure. Its implementation was salvaged into #68153, whose diff also handles the resulting k3/kimi-k3 duplicate-row case.
• #68153 [merged] duplicate — (+354/-12) — merged reference implementation: Preserves #67409’s cross-picker search-alias fix and adds canonical alias-based merge deduplication so curated kimi-k3 and live k3 do not appear as two rows; it remains relevant because it is the merged implementation that supersedes #67409.

Duplicates

#67409 and #68153 implement substantially the same cross-picker k3 ↔ kimi search-alias change; #68153 is the superseding version with additional alias-aware deduplication.

Suggested consolidation

Close #67409 as duplicate of #68153: #68153 explicitly salvages its picker-search fix, includes the same affected surfaces and tests, and extends it with regression coverage and canonical deduplication for k3 versus kimi-k3. The duplicate chain is #67409 → #68153, with #68153 already serving as the merged reference implementation.

Complex graph

flowchart LR
    classDef open fill:#dbeafe,stroke:#1d4ed8,color:#1e3a8a
    classDef merged fill:#dcfce7,stroke:#15803d,color:#14532d
    classDef closed fill:#e5e7eb,stroke:#6b7280,color:#1f2937
    classDef unverified fill:#f3f4f6,stroke:#9ca3af,color:#374151
    classDef best stroke-width:3px,stroke:#b45309
    classDef target stroke-width:3px,stroke:#4338ca
    subgraph Dup67409 ["PRs duplicating each other"]
        P67409["PR #67409 (open)"]
        P68153["PR #68153 (merged)"]
    end
    class P67409 open
    class P68153 merged
    class P67409 target
    click P67409 "https://github.com/NousResearch/hermes-agent/pull/67409"
    click P68153 "https://github.com/NousResearch/hermes-agent/pull/68153"

Loading

Graph: solid arrow = fixes / best fix, dashed arrow = partial or unverified (see edge label); boxed group = PRs duplicating each other; amber border = best fix; indigo border = target; gray node = closed (state tag in the node label).

Cross-PR triage: Reviewed 2 pull requests and 0 issues in this complex. Each diff was read against this issue; Assessment working set: 37 kB of PR diffs, 3 kB of issue/PR text, <1 kB of discussion (2 comments), 0 verify verdicts. verdicts reflect diff content, not PR titles. Part of an automated triage batch.