Commit Graph
306 Commits
Author SHA1 Message Date
ciregenz 685314b313 [eric] free-trial: skip the idle cache-warm ping on the free lane (useless across rotating pool accounts, and it'd burn a run at idle) 2026-06-13 19:06:57 -07:00
ciregenz bf8b450e33 [eric] free-trial: tag the remaining per-query aux calls (turn-label, group-meta, cache-warm, preflight) so one query stays one run 2026-06-13 19:03:35 -07:00
ciregenz 846c13e8ad [eric] free-trial: tag title-gen with the session id so a query + its title-gen = one run 2026-06-13 18:52:51 -07:00
ciregenz 01df3cde17 [eric] free-trial: fresh task-id per query so each query is its own run (cancel or not) 2026-06-13 18:42:49 -07:00
ciregenz e4740b74c1 [eric] free-trial: pin the trial to Haiku end-to-end (desktop + subagent) so the CLI never sends effort 2026-06-13 17:38:06 -07:00
ciregenz c8a86cd3ff [eric] free-trial: set ENABLE_TOOL_SEARCH=auto on the proxy branch (tool-defer+cache_control 400) 2026-06-13 17:17:28 -07:00
ciregenz b9d90250b7 [eric] updater: evergreen silent install-on-idle (staged update + idle + no agents) 2026-06-13 16:06:26 -07:00
ciregenz 8b6a65fc4b [eric] free-trial: tag free-lane calls with the session id so 1 task = 1 run 2026-06-13 16:03:02 -07:00
ciregenz 5291a4ae60 [aidan] merge #79: QoL, sidebar ui/ux, titles, text compaction, bug fixes 2026-06-13 02:06:54 -07:00
ciregenz 490bbf4113 [aidan] merge #80: cross-origin iframe perception + spawned card end UX 2026-06-13 01:53:47 -07:00
AidanandGitHub 3f4d91d8b4 [aidan] bug: cross-origin iframe perception + spawned card end UX (#78)
Browser sub-agents can now drive Google Drive’s share-doc flow end to end: same-process cross-origin iframes (the share dialog) are visible through a frame-tree-aware AX walk, and combobox autocomplete fields fill in one click_index call via DOM.focus + Input.insertText. When a spawned browser card’s agent finishes, the card now fades through a “Task done” pill with a Keep button for ~3s instead of snapping out, so the user sees what happened.
2026-06-13 00:58:44 -07:00
Aidan 7b88be7003 [aidan] ui/ux: hide noisy memory/mode from chat title, fix token bug
Drop the inline 'Memory N% full' pill from AgentChat and the mode label from AgentCard so the chat header stays quiet until tools are connected. Backend now emits agent:context_update after compact and clear so the session token counters reflect the trimmed history instead of stale pre-compact totals, with a new _estimate_post_compact_input helper and matching invariants in test_v2_invariants.
2026-06-13 00:40:19 -07:00
AidanandGitHub 08526b6149 [aidan] bug/feat: sidebar reordering, title improvements, openai gen bug (#76)
* [aidan] ui: sidebar reorder

* [aidan] ux: sidebar naming bug fix

* [aidan] ui: title changing improvements

* [aidan] ui/ux: aux title gen streams for cx/ route + agent writes meta.json first

- All aux LLM calls (chat title, turn label, group meta, dashboard name) now stream
  instead of using messages.create; 9router's cx/ subscription response translator
  drops content for GPT-5-family models on the non-streaming path but works per-event.
- aux_max_tokens_for floors GPT-5 budget at 4K so reasoning still leaves headroom for
  a label; non-reasoning models get the base 100.
- App Builder skill: write meta.json FIRST (step 1 of Quick start) so the app's name
  surfaces in the sidebar + ViewEditor header on the agent's first tool call instead
  of waiting until end-of-turn.
- Session-end sync_output_from_meta_json takes a fallback_name (= session.name) so an
  agent that never writes meta.json still leaves the app with the aux-LLM chat title
  rather than "Untitled App".
- Title display: truncateForTitle caps at 4 words / 30 chars; displayChatTitle picks
  the right Phase 1 placeholder by session.mode; normalizeSessionName strips legacy
  Agent-XXXX names at slice intake.
- Typewriter component (char-by-char delete-then-type) drives the chat header,
  dashboard card title, sidebar Apps entry, and ViewEditor TextField + description
  field; honors useReducedMotion.

* [aidan] tune: reduce gpt-5 aux token floor
2026-06-12 23:33:32 -07:00
ciregenz a1f390ef0f [eric] browser: claim only the user-picked card, not every dashboard card (fix 5cf6e2a) 2026-06-12 20:06:56 -07:00
ciregenz 5cf6e2a4b1 [eric] browser: honor the user's selected browser card instead of spawning a new one (selection now reaches sub-agent dispatch + is claimed for the task) 2026-06-12 17:31:46 -07:00
ciregenz 0a81bfa17a Merge PR #70 [Aidan] appbuilder: inject selected-app context into dashboard agent chat 2026-06-11 02:51:05 -07:00
ciregenz ddd0641cac [eric] merge free-trial + value-first onboarding (flag off, reorder gated so no-trial = today's flow)
# Conflicts:
#	frontend/src/app/pages/AgentChat/bubbles/MessageBubble.tsx
2026-06-10 23:20:18 -07:00
ciregenz 6b5f10e287 [eric] bash defaults to always-allow like other actions (catastrophic/scheduling/sensitive-path guards still prompt); drop the working status dot 2026-06-10 23:13:02 -07:00
AidanandGitHub e8e19661e6 Inject selected-app context into dashboard agent chat (#69)
When a user selects an App card on the dashboard via the "Select UI element" picker, thread the selected Output ids to the backend (mirroring the selected_browser_ids path) and inject a per-app context block into the system prompt: each app's absolute workspace path, entry point, meta.json, and a pointer to its SKILL.md. This lets the agent edit existing apps in place from the main view (the dashboard card's Vite runtime live-reloads on save) without switching to the Apps tab.
2026-06-10 21:29:02 -07:00
ciregenz 100d6491f7 [eric] models: track model errors (diagnostic) + friendly 'not on your plan' card for the 1211 unknown-model wall 2026-06-10 19:10:57 -07:00
ciregenz a126c42a07 [eric] free-trial: hand the wheel back when a real model is connected mid-trial (key or 9router sub) 2026-06-10 16:32:25 -07:00
ciregenz fb178648df [eric] free-trial: backend dispatch injection, hardware fingerprint, mint/clear, exhaustion upsell 2026-06-09 23:54:10 -07:00
Aidanandciregenz 0a704d4690 [aidan] chat: isolate branch history, fresh session on edit/switch (#62) 2026-06-09 22:29:31 -07:00
ciregenz 1fd32b7bc3 [eric] backend: add Fable 5 subscription (cc/) row, verified live 200 on the Claude sub 2026-06-09 15:58:19 -07:00
ciregenz 7ab61057df [eric] backend: surface Fable 5 in the model picker on the api-key route + price it $10/$50 2026-06-09 14:55:07 -07:00
ciregenz 1be1ce581d [eric] providers: Fable 5 turns thinking off by omitting the param, not thinking:disabled (which 400s) 2026-06-09 14:06:18 -07:00
ciregenz a774f68b96 [eric] providers: add Claude Fable 5 on the API-key route (sub route held on pinned 9router) 2026-06-09 13:27:45 -07:00
ciregenz e24a63c13c [eric] browser: steer paginated gathers to sweep in batch (RepeatFlow + sessionStorage / scroll-then-read), not one page per turn 2026-06-09 02:39:31 -07:00
ciregenz bfcad98b76 [eric] browser: BrowserSaveData writes a big page dataset to one sandboxed file, ending the dispatch-per-chunk waste 2026-06-09 02:05:26 -07:00
ciregenz 9eb0e2cb2b [eric] browser: new gathered data resets the spin backstop, so a page-by-page gather runs to completion instead of getting cut with partial results 2026-06-08 23:47:26 -07:00
ciregenz ba41409577 [eric] browser: spin backstop nudges a clean Done wrap-up, not a mid-thought cutoff 2026-06-08 23:26:15 -07:00
ciregenz 438d1e36d4 [eric] browser: arm the send-completion shortcut only for send tasks, not gathers (cookie click faked a send) 2026-06-08 23:02:55 -07:00
ciregenz 07f911d200 [eric] browser: a clean success needs an explicit Done; recognize gather asks, steer them to BrowserExtract 2026-06-08 22:37:33 -07:00
ciregenz 91073becec [eric] browser: finish via a Done(message) tool so the user gets a plain conversational reply, not the OUTCOME tag 2026-06-08 22:15:52 -07:00
ciregenz 110dc1c745 [eric] browser: fast-path emits a synthetic BrowserAgent tool-call/result so the 'Browser Agent' bubble always shows, even when the orchestrator is skipped 2026-06-08 13:07:12 -07:00
ciregenz b634638677 [eric] browser: poll the interactives list for the real Send button after a composer fill so the handoff actually fires (was 0/10, racing the lazy render) 2026-06-08 12:42:18 -07:00
ciregenz 8d9c75803c [eric] browser: restore the Send-button handoff (points the model at the Send index after a fill); deleting it brought the Send-hunt back, it was NOT redundant with the wait 2026-06-08 00:58:38 -07:00
ciregenz bf5f95573a [eric] browser: wait up to 6s (not 2.5s) for the lazy-rendering Send button after a composer fill, kills the re-list turn and the late-Send miss 2026-06-08 00:44:26 -07:00
ciregenz cdba825277 [eric] browser: final message is ONLY the OUTCOME line in plain human words, no narration preamble or UI jargon 2026-06-08 00:31:27 -07:00
ciregenz 2d59a0fbf1 [eric] browser: stall-path completion text is a plain confirmation, never the raw action-log proof (indices/coords) 2026-06-08 00:15:39 -07:00
ciregenz f8c812353e [eric] browser: agent writes a plain human final message (no UI jargon/OUTCOME mechanics); delete the leaky send-ready advisory, dead prefix flag, and humanize aux experiment 2026-06-08 00:09:50 -07:00
ciregenz ce8e74b96a [eric] browser: skill-replay returns a normal OUTCOME line so the parent translates it, no more 'learned skill replay (N steps, no LLM)' leaking to users 2026-06-07 23:49:21 -07:00
ciregenz 725578b2a8 [eric] browser: orchestrator translates the OUTCOME line into a plain human confirmation; drop hardcoded task-specific examples from prompts 2026-06-07 23:44:39 -07:00
ciregenz d5fc146123 [eric] browser: never leave a tool_use unanswered (backfill mid-turn + validate both directions) so a 30s upstream reset stops 400-killing the run 2026-06-07 23:04:59 -07:00
ciregenz 2409e2d4cd [eric] browser: fire the learning self-audit every 25 finished tasks in a daemon thread (zero hot-path latency) 2026-06-07 21:33:38 -07:00
ciregenz 6ef7b33667 [eric] browser: cross-site meta-playbook (tier 3) + self-audit of the learning machinery (tier 4, proposes for human review, never self-edits) 2026-06-07 21:20:40 -07:00
ciregenz 7ec05613fe [eric] browser: hand the Send button to the model after a composer fill, and commit to an already-open thread instead of re-verifying 2026-06-07 19:28:51 -07:00
ciregenz 1204e733b5 [eric] browser: stall backstop only fires after a real action, never cuts short early orientation 2026-06-07 19:12:43 -07:00
ciregenz e19def1ae5 [eric] browser: a clean Send-class click ends the run; bound the look-only spin even when the send cant be auto-detected 2026-06-07 19:07:34 -07:00
ciregenz 0f63697765 [eric] browser: confirm waits for the target to actually render so a send isnt falsely 'not confirmed' under load 2026-06-07 18:58:43 -07:00