Commit Graph
24 Commits
Author SHA1 Message Date
ciregenz f59bf0db9b defluff (frontend + backend): strip em-dashes + shorten docstrings + drop dead UI files (cosmetic only, no schedule code) 2026-05-20 05:36:17 -07:00
ciregenz a1fefa0ae8 [eric] approvals: trust patterns for action approvals + smarter sensitive-file gating 2026-05-20 01:56:38 -07:00
ciregenz d20ff7d16c [eric] app builder cleanup — remove Auto Run, Test Input, RenderOutput tool + Outputs activation gate 2026-05-11 09:05:45 -07:00
ciregenz fae2806124 [eric] make non-Anthropic API keys actually work end-to-end, fix the wildly wrong cost numbers, multi model ux/ui improvements 2026-05-06 21:05:57 -07:00
ciregenz c64bf65c66 [eric] session state additions + ui hook refinements 2026-05-05 17:05:54 -07:00
ciregenz 59e20e11ea [eric] clean up remaining non-opaque references 2026-05-04 22:25:48 -07:00
ciregenz b7a1faee38 [eric] delete miscelaneous scrapped code 2026-05-04 19:15:42 -07:00
ciregenz 8286cc1a5c [eric] delete ~3.6k lines of dead code (old unused agent runtime, ghost fields, orphan scripts). all 660 tests still pass, no functional
change.
2026-05-03 22:39:17 -07:00
ciregenz d14cc89300 [eric] tool labels read like a person — Plugged into Gmail, Saved a snapshot, varied phrasings per call; session/turn labels
in sentence case; activating an MCP mid-chat now actually works instead of the model guessing at made-up tool names. Also a bunch miscelaneous ui/ux tweaks, I cant be bothered :)
2026-05-03 20:44:32 -07:00
ciregenz ade79315a9 [eric] thinking pill shows full turn cost (input + output + subagent + tool work) with click-to-see input/output breakdown; reopening
completed chats no longer replays the typewriter (per-session lastSeq survives AgentChat remount so resume protocol stays at the high-water
  mark instead of last_seq=0)
2026-05-03 11:21:02 -07:00
ciregenz d310f9e031 [eric] thinking pill works on GPT/Gemini (clickable bubble with duration + token count + honest "provider does not expose
reasoning text" note instead of vanishing); friendlier MCP tool names (Gmail/Slack/etc with verb-derived actions); auto-collapse Anthropic
  thinking pill on turn end so the answer comes first; header timer matches per-turn pill (sums in ms, rounds once); thinking pill rolls 251s
  to 4m 11s; parallel tool count chips do not jitter as N grows; drop Gemini 2.5 from picker (Gemini 3 family only); aux-LLM calls route
  per-model so cheap-tier auxiliary calls do not 401 on cc/ subscription routes
2026-05-02 23:46:13 -07:00
ciregenz 691a92c2a1 [eric] dynamic turn labels in the thinking pill via cheap-tier aux LLM + cache pre-warm on dashboard mount for faster turn-1
TTFT
2026-04-30 12:31:48 -04:00
ciregenz 98349df9f1 [eric] faster replies + smarter chat: optimistic message bubbles, live thinking... pill, friendlier tool-call labels (Reading → Read),
compaction chip, native completion notifications, prompt-cache flip for ~70% cheaper/faster
2026-04-29 14:07:18 -04:00
ciregenz 5d69215739 [eric] phase 2+3: outputs gate, smart compaction, slash commands, preflight + auto-continue, bumped Anthropic SDK 2026-04-28 19:55:06 -04:00
ciregenz 30d581e9ec [eric] phase 1: MCP activation gate, context pill, friendly 429s, -api routes so API keys actually work 2026-04-28 19:01:07 -04:00
ciregenz 14e1d02bbb [eric] reasoning UI + thinking-level controls: collapsible 'Thought for Ns' bubble with shimmer, per-session thinking level (off/low/med/high/auto) on
reasoning-capable models, gate GitHub Copilot as 'Coming soon', delete Twitter/xbird integration, fix model-picker first-click + 200K context cap on Sonnet/Opus, revive
  arrow-key dashboard nav after typing/clicking/zooming
2026-04-15 12:47:34 -07:00
ciregenz 4778cea80e [eric] multi-model subscription support: connect ChatGPT Plus, Gemini Advanced, and GitHub Copilot subscriptions via 9Router
- Add BUILTIN_MODELS for OpenAI (GPT-5.4/Mini/5.3-Codex), Google (Gemini 3 Pro/Flash, 2.5 Pro/Flash), and GitHub Copilot
  - Add resolve_model_id_for_sdk() and resolve_aux_model() to route prefixed model IDs (cx/, gc/, gh/) through 9Router translator
  - Add GET /agents/models endpoint returning available models based on live 9Router connection state
  - Enable ENABLE_TOOL_SEARCH=auto for all providers so non-Claude models get full tool access (23 built-in + all MCPs)
  - Add thinking block streaming support (ThinkingBlock in stream handler + AssistantMessage) for reasoning models
  - Fix Gemini 3 thought-signature errors: use skip_thought_signature_validator per Google docs
  - Fix WebSearch blocked_domains/allowed_domains empty-list rejection on Anthropic API
  - Fix browser_agent text_parts UnboundLocalError on non-Claude models
  - Fix auxiliary LLM calls (title gen, group meta, dashboard naming, view builder, browser agent) via resolve_aux_model
  - Add Codex OAuth callback listener on port 1455 for ChatGPT Plus subscription connect
  - Route Gemini OAuth through system browser since Google blocks embedded webviews
  - Override Electron popup user-agent for OAuth
  - Fix duplicate OAuth callback with idempotent completed_oauth tracking
  - Fix Settings modal tab routing and stale warning banner
  - Add API key inputs for OpenAI, Google, and OpenRouter
  - Add MCP warning banner when selecting non-Claude model with many tools
  - Force session fork on cross-provider model switch to prevent transcript corruption
  - Disable GitHub Copilot subscription card (9Router poll issue, marked preview)
  - Delete dead CopilotProvider import and unused CopilotAuthButton component
2026-04-12 12:06:36 -07:00
ciregenz 75d5d8c382 [eric] fix context loss on edit, branch, and resume: fork SDK session instead of clearing it, load stopped sessions from disk 2026-04-01 18:20:19 -07:00
ciregenz ebea25f0c2 [eric] [eric] add PostHog analytics, 9Router subscription proxy, unified usage tracking
- PostHog integration: collector, analytics subapp, opt-in UI, Analytics page
  - 9Router: auto-start, OAuth subscription flow, /v1/messages Anthropic format support
  - Settings overhaul: multi-provider API keys, subscription connect UI, onboarding modal
  - Unified usage: merge 9Router cost/token data into Settings Usage tab
  - Provider system: providers/, agent_loop, tools/ (unused, for future non-Anthropic support)
  - Agent SDK: restored as primary with 9Router ANTHROPIC_BASE_URL fallback
  - Updated system prompt, credential resolution, dashboard analytics
2026-03-24 14:02:58 -07:00
haikdc a1138a1782 [Haik]: (ability to resume chats if force end early) (agent selection actually works as advertised) (browser control/creation tool isolation) (in dev mode can see the action input schema) 2026-03-18 14:19:42 -07:00
haikdc 6f02daff2d [Haik]: (Browser actions shown in agent card output) (User can have item selection turned on by default in settings) (most browser related functionality works - altho spotify is tricky) (mcp discovery in prod shld worke now?) 2026-03-17 01:48:50 -07:00
haikdc b90e5c6353 [Haik]: (Practically all forms of dashboard interaction glitches) (made dashboard selection process much more intuitive and seamless - and high level) (added BrowserAgent so the outer agent can delegate more complex tasks) (added tool to parallelize across multiple browser agents since the abstraction keeps the tool calls so high level) (Added a temporary agent insight ui over each browser as it does stuff - human in the loop is still needed) 2026-03-16 05:01:56 -07:00
haikdc 1632b5c9e5 [Haik]: ckpt, (Fixed view create duplication bug) (Views auto save) (MCP is star sorted by default) (reload issues in dash fixed) (electron app made) 2026-03-15 00:09:04 -07:00
haikdc 23621306a1 [Haik]: and so it begins 2026-03-13 12:09:25 -07:00