* feat(anthropic): add Claude Opus 5 model support
Wire claude-opus-5 through the Claude 5 contract seams: adaptive-by-default
thinking with the full low..max effort range, default-sampling and prefill
stripping, streaming refusal contract, model-bound thinking replay, and 1M/128k
catalog metadata across anthropic, claude-cli, Vertex, Bedrock, and Mantle.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(anthropic-vertex): normalize explicit Vertex Opus 5 model rows
Extend normalizeAnthropicVertexResolvedModel so user-configured Vertex Opus 5
rows regain reasoning, image input, 1M/128k limits, and the native thinking
map; update the live-model priority expectation and regenerate docs_map.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
* fix(deepinfra): tag DeepSeek thinking format and use OpenAI video endpoint
Route DeepInfra video generation through the OpenAI-compatible async
/v1/openai/videos endpoint (submit, then poll) with live video-gen tag
discovery, replacing the retired native /v1/inference path; doctor
migrates api.deepinfra.com URLs and the provider fails closed with an
actionable error on retired custom native endpoints without echoing
configured URLs (they may carry credentials).
Tag the deepseek-ai/* family with thinkingFormat "deepseek" so core
strips/recovers reasoning_content and DSML tool-call markup instead of
leaking it into replies, and mark reasoning-tagged models as
reasoning-capable.
Redaction tests assemble credential-shaped fixture URLs from string
pieces so TruffleHog's URI detector does not flag them. Docs describe
the per-surface static defaults with live discovery expanding the
selectable model lists.
* fix(deepinfra): preserve custom video output host
---------
Co-authored-by: Jacob Tomlinson <jacobtomlinson@users.noreply.github.com>
* fix(anthropic): accept "cli" entrypoint in Claude session catalog discovery
Claude Code v2.x writes entrypoint: "cli" in its JSONL session files.
Previously only "sdk-cli" was recognized, causing all v2.x sessions to
be silently skipped.
Extract CLI_ENTRYPOINTS set and isCliEntrypoint() helper, then apply
the same fix to both TypeScript catalog and macOS paired-node native
catalog. Add matching regression coverage on both platforms.
Closes#105164
* fix(anthropic): refresh Claude CLI session discovery
Apply the reviewed cli/sdk-cli allowlist, mirrored negative coverage, and fallback-scope documentation to current main.
Co-authored-by: 黄攀0668000858 <huang.pan@xydigit.com>
---------
Co-authored-by: Peter Steinberger <steipete@gmail.com>
The live codex text provider was a redundant projection of the openai
catalog (exclusive provider ownership; the openai plugin's ChatGPT OAuth
discovery already serves gpt-5.6-* route-aware). Folding it:
- extensions/codex no longer registers a text provider, catalog entry, or
synthetic text auth; provider.ts/provider-catalog.ts/provider-discovery.ts
and the route-blind model-name heuristics are deleted; the narrow
post-harness reasoning fallback moves to an app-server-owned module
- openai thinking policy keys on explicit selected-route provenance
(api === openai-chatgpt-responses) instead of value-shape inference
- models.list gains an optional additive agentRuntime field (configured
intent); session agentHarnessId remains the execution proof
- doctor --fix migrates the shipped codex/* config shape end to end:
every model slot, provider-config merge with blocker-aware conflict
handling, sessions, cron payloads (two-phase: runtime policy persists
before cron refs rewrite), transcripts; migrated refs carry model-scoped
agentRuntime.id=codex preserving the shipped wizard semantics; auto
runtime policies normalize to codex with sibling fields preserved;
blocked provider conflicts retain the whole legacy namespace fail-closed
with an actionable warning
- the stale openai:default profile cleanup (#91352) was deliberately
deferred to a follow-up after review showed it needs per-agent identity
proofs; doctor keeps warning about unusable profiles
Fixes#105561Fixes#84637Fixes#90420
* fix(deepgram): reject malformed and non-http(s) realtime base URL overrides
* fix(deepgram): accept direct ws(s):// realtime base URL overrides
Preserve released behavior: v2026.6.11 passes wss:// (and ws://) base URL
overrides straight through to the WebSocket URL builder. The prior validator
rejected every non-http(s) scheme, breaking custom Deepgram realtime proxies
that use a direct wss:// endpoint. Accept ws:/wss: alongside http(s):, keep
rejecting malformed and unrelated schemes, and preserve a direct ws(s):
override's protocol/path/port through toDeepgramRealtimeWsUrl.
* fix(deepgram): keep secure ws:// -> wss:// upgrade for realtime base URL
A prior change preserved direct ws:/wss: overrides unchanged, which also
stopped upgrading a plaintext ws:// override to wss://. That regressed the
released behavior where every accepted non-http: scheme maps to wss:, so an
existing ws:// endpoint would begin sending Deepgram realtime audio and
provider auth over an unencrypted socket. Restore the release mapping
(http: -> ws:, all others -> wss:), which still preserves a direct wss://
override and keeps the ws:// -> wss:// secure upgrade.
* fix(deepgram): preserve custom websocket transport
Co-authored-by: dwc1997 <du.wenchi@xydigit.com>
* test(deepgram): use placeholder credential fixture
* test(deepgram): clarify loopback callbacks
---------
Co-authored-by: Peter Steinberger <steipete@gmail.com>
* feat(openai): recognize GPT-Live realtime models and fail closed with guidance
OpenAI's gpt-live-1/gpt-live-1-mini full-duplex voice models are API
early-access only and use a WebRTC-only quicksilver session protocol with
handoff-based agent delegation that OpenClaw's realtime transports do not
implement yet. Configuring a gpt-live-* model now produces actionable
configuration errors on the realtime WebSocket bridge and Talk browser
sessions instead of opaque provider errors or audio-only sessions without
agent access. GA gpt-realtime behavior unchanged; default model stays
gpt-realtime-2.1.
Related: #104683
* test(openai): narrow browser session union before asserting offerUrl
The Codex/Claude sidebar sections leaked long titles across the row
timestamps and past the sidebar edge because the row body was never a
flex container, so the title span stayed inline and ellipsis rules did
not apply. The sections also rendered every loaded session (Codex swept
up to 100 catalog pages per host every 30s) with no height cap, growing
past the sidebar shell and painting over the footer.
Rows now reuse the regular session-row __text anatomy so titles
ellipsize, the sidebar renders only the newest 10 sessions per host
with a derived truncation note (the full catalog stays on the sessions
page), each section shrinks and scrolls inside the sidebar body, and
the shell body clips as a final guard. The sidebar full-catalog
hydration loop and the one-off __body class are deleted.
Formatting proof: oxfmt --check clean on changed files via Testbox
tbx_01kx92cwpr1x9ccp2fgn441y8h (local hook skipped: no node_modules).
Fixes#104584
* fix(opencode-go): remove deprecated mimo-v2-omni and mimo-v2-pro model aliases
These deprecated aliases reject agent requests from the OpenCode Go gateway.
Remove them from the provider catalog and clean up all references in probe
skip lists, CI workflows, and tests.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* fix(opencode-go): complete deprecated MiMo cleanup
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-authored-by: Peter Steinberger <steipete@gmail.com>
