Commit Graph
273 Commits
Author SHA1 Message Date
ciregenz bcfbb8d24f [eric] browser: fail fast with an honest message when the card's webview is gone 2026-06-02 18:19:33 -07:00
ciregenz 34faca12ae [eric] browser: resolve the live card host for replay and report did-nothing runs honestly 2026-06-02 18:14:57 -07:00
ciregenz 68017465ce [eric] browser: add a deterministic completion-honesty gate to catch ghost successes 2026-06-02 18:14:57 -07:00
ciregenz 2b2d5909a8 [eric] browser: instrument skill lifecycle, measure replay speedup, flag silent non-help 2026-06-02 18:03:43 -07:00
ciregenz 294e96d782 [eric] browser: version skills behind a trust gate, compose on proven sub-skills 2026-06-02 18:03:36 -07:00
ciregenz 5bb0b46890 [eric] browser: give the agent tools to list and deprecate its own learned skills 2026-06-02 16:33:21 -07:00
ciregenz e2a77b5eae [eric] browser: parameterize skills so one skill serves a task with different inputs 2026-06-02 16:27:11 -07:00
ciregenz e30d70ce97 [eric] browser: persist skills across sessions with secret-redaction gate 2026-06-02 16:17:23 -07:00
ciregenz 3aa9cf6ba0 [eric] browser: front-load page perception into turn 1 to cut orientation round-trips 2026-06-02 14:02:47 -07:00
ciregenz a015036383 [eric] browser: prompt-cache the fixed prefix and push batching to cut first-run turns 2026-06-02 14:01:00 -07:00
ciregenz 014414440b [eric] browser: decompose BrowserBatch into discrete replayable skill steps 2026-06-02 13:32:54 -07:00
ciregenz d1c1b44ee7 [eric] browser: replay learned skills with zero LLM calls, fall back on any miss 2026-06-02 12:49:55 -07:00
ciregenz d6682190cc [eric] browser: in-memory skill cache that distills tasks into replayable steps 2026-06-02 12:45:42 -07:00
ciregenz 69c4546735 [eric] browser: refine ghost detector so read tasks need returned content, not state change 2026-06-02 12:35:12 -07:00
ciregenz a8964830ac [eric] browser: record per-tool latency, tokens, and per-task completion metrics 2026-06-02 12:12:41 -07:00
ciregenz ea823951c6 [eric] browser: loop test for WebMCP-detect and route list/replay tools 2026-06-02 04:15:08 -07:00
ciregenz 15781d2a3b [eric] browser: list captured routes and replay safe GETs to skip the UI (tier 2) 2026-06-02 04:11:50 -07:00
ciregenz 6e006dd472 [eric] browser: detect a page's declared WebMCP tools (tier 1) 2026-06-02 04:03:10 -07:00
ciregenz 43ba7d2f1c [eric] browser: unit and end-to-end loop tests for the new browser behaviors 2026-06-02 03:56:07 -07:00
ciregenz 82dcb6bb4e [eric] browser: per-domain hints, deterministic stagnation, and aux-LLM fallback 2026-06-02 03:56:07 -07:00
ciregenz 63e321b9fa [eric] agent: call out the literal "-" in the no-dashes voice rule 2026-06-01 15:48:23 -07:00
ciregenz c89ba4dc30 [eric] agent: give the assistant a tight, quirky, no-dashes texting voice 2026-06-01 15:01:13 -07:00
ciregenz 3d65cacddf [eric] tools: drop personal gmail from the connected-account comment example 2026-06-01 15:01:13 -07:00
eric a3114367d9 [eric] merge: streaming cadence + live markdown + auth/devtools/browser fixes from dev 2026-06-01 11:55:38 -07:00
eric bcb88ebefb [eric] auth: dev-only /api/dev/token so the split-port frontend can authenticate (404s when packaged) 2026-06-01 11:55:17 -07:00
eric 5fdbc3ab79 [eric] router: revert dev pin to 0.3.60 (0.4.x adds an api auth gate that 401s our oauth/provider-sync calls, breaking every subscription connect); documented what porting the bump needs 2026-06-01 08:27:00 -07:00
eric f40d3b3b95 [eric] router: bump dev 9router pin 0.3.60 -> 0.4.66 to test websearch rework + opus-4-8/gpt-5.5 sub routes (prod pin stays 0.3.60 until websearch verified live) 2026-06-01 08:12:35 -07:00
eric 4b3aa3eb8b [eric] fix: revert enable_tool_search back to auto (forcing 1 marked tools defer_loading and 400'd the claude sub route on the defer_loading+cache_control combo) 2026-06-01 07:58:09 -07:00
eric 1a4a23f131 [eric] models: gemini 3.5 flash is api-key only since pinned 9router 0.3.60 has no gc/ route for it (verified against the 0.3.60 model registry); opus 4.8 cc/ confirmed via passthrough 2026-06-01 00:03:13 -07:00
eric 581503d1d3 [eric] models: drop gpt-5.3-codex + high/xhigh (superseded by gpt-5.5, high/xhigh were just effort variants not real models) 2026-05-31 23:59:59 -07:00
eric aff1b5f5bb [eric] models: add claude opus 4.8 (new flagship) + gemini 3.5 flash (GA), drop dead gemini-3-pro (shut down mar 9), verified against official anthropic/openai/google docs 2026-05-31 23:53:31 -07:00
eric 8d93e29321 [eric] agent: reframe identity as the user's personal digital company, drop coding emphasis 2026-05-31 23:38:53 -07:00
eric 979301db4d [eric] agent: trim identity blurb to one OpenSwarm mention and a tighter line (79 tokens) 2026-05-31 23:37:26 -07:00
eric 5e4148ef3e [eric] agent: make the identity blurb shorter, quirkier and less repetitive (mistaken-identity framing, one disavowal not three) 2026-05-31 23:33:13 -07:00
eric 1bccb24b54 [eric] agent: stop the assistant calling itself Claude Code by appending an identity override after the preset (keeps tools + cache, edit AGENT_NAME to rebrand) 2026-05-31 23:29:18 -07:00
eric 11b7d9b7a4 [eric] perf+ui: shrink runs chunks in parallel (5min -> ~1min) and shows a live pulsing label so users know its working 2026-05-31 18:50:46 -07:00
eric 56caabebbc [eric] fix: popup uses full width and matches app surface, error toast no longer red wall, summarize chunks small enough for dense pdf tokenization 2026-05-31 18:40:33 -07:00
eric 1372130d28 [eric] fix: summarize big files by chunking when source exceeds aux model window then merging 2026-05-31 18:33:02 -07:00
eric 3dc8ff842a [eric] fix: text attachment cap is all-or-nothing now, no half-files since partial confuses the model and the user 2026-05-31 18:24:38 -07:00
eric 0ed5457ca2 [eric] docs: record hardening precedences (compaction trims, ssrf is async loopback-allowed, ports must be stable) 2026-05-31 18:13:06 -07:00
eric 1035c50be3 [eric] perf: force enable_tool_search=1 on non-anthropic paths so schemas load on demand and first message is ~9k tokens cheaper 2026-05-31 18:06:39 -07:00
eric 0b35b70d0b [eric] sec: wire ssrf guard into webfetch (per-redirect) and /api/web/fetch (initial) 2026-05-31 18:05:34 -07:00
eric 68b7aa9a94 [eric] sec: ssrf guard module (async dns, ipv6 covered, loopback allowed for app builder) 2026-05-31 18:04:16 -07:00
eric c9d2005e0b [eric] fix: cap combined inline text attachments at 1.5m chars so a pile of txt files cant silently blow the context window 2026-05-31 18:01:48 -07:00
eric ecf348f9cb [eric] fix: autocompact now actually trims the model's history instead of just showing a marker 2026-05-31 18:00:39 -07:00
eric 72e0bf4105 [eric] fix: vendor kirbah-mcp-youtube.js bundle (33mb, mostly googleapis) to skip npx on first run 2026-05-31 16:07:53 -07:00
ciregenz 4c6326a566 [eric] security: pin 9router to loopback so it stops lan-exposing api keys + tripping the firewall prompt 2026-05-29 22:17:26 -07:00
Eric 41031a7790 [eric] backend: load_settings never bricks boot on a corrupt or version-drifted settings.json - preserve unparseable files aside and drop only the invalid fields instead of raising, plus a migration test suite 2026-05-28 21:41:01 -07:00
Eric e8fd9ceb5a [eric] backend: add preflight_enabled (default true) and preflight_rollout_pct (default 100) to AppSettings so the runtime preflight can be killed per-install via the existing settings sync or staged 5/25/100 by the cloud cohort gate without a release 2026-05-28 02:49:06 -07:00
Eric e8f32da42d [eric] backend: pin the six previously-unpinned requirements (jsonschema, fastapi, pillow, httpx, trafilatura, tzlocal) to the exact versions baked into the last working packaged build so deps-pinned passes and reproducibility actually matches the header comment 2026-05-28 00:55:55 -07:00