readme rewritten with ascii diagrams, detailed setup, and conversational tone.
root package.json updated with packageManager field for bun, web workspace,
and build/start scripts pointing to web/ for railway railpack compatibility.
Skills (copy_visible_text, find_and_tap, submit_message, read_screen,
wait_for_content, compose_email) were CLI-only using direct ADB. The
server prompt advertised them but they silently failed when chosen.
Now intercepted in the agent loop before actionToCommand() and executed
server-side using existing WebSocket primitives (get_screen, tap, swipe,
clipboard_set). Each skill replaces 3-8 LLM calls with deterministic
server-side logic.
The UI agent had no memory of previous actions — each step was a fresh
single-shot LLM call. After typing and sending a message, the LLM saw
an empty text field and retyped the message in a loop.
- Add RECENT_ACTIONS (last 5 actions with text/result) to user prompt
- Add chat app completion detection rule to dynamic prompt
- Add send-success hints for WhatsApp and Messages apps
- Add git convention to CLAUDE.md (no co-author lines)
- Add empty goal guard in parser (returns done instead of passthrough)
- Replace `as any` casts in pipeline.ts with proper ActionDecision types
- Add runtime type guards for untrusted LLM output in classifier
- Add intent action to dynamic prompt so UI agent can fire intents
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replace preprocessor+runAgentLoop with runPipeline in both device.ts
(WebSocket) and goals.ts (REST). The pipeline orchestrates: deterministic
parser (stage 1) -> LLM classifier (stage 2) -> lean UI agent (stage 3).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
When pipelineMode is enabled in AgentLoopOptions, the loop uses
buildDynamicPrompt() with per-screen context (editable fields,
scrollable elements, app hints, stuck state) instead of the static
mega-prompt. Legacy mode (default) is unchanged.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Android: fetch installed apps via PackageManager, send to server on connect
- Android: add QUERY_ALL_PACKAGES permission for full app visibility
- Android: fix duplicate Intent import, increase accessibility retry window
- Android: default server URL to ws:// instead of wss://
- Server: store installed apps in device metadata JSONB
- Server: inject installed apps context into LLM prompt
- Server: preprocessor resolves app names from device's actual installed apps
- Server: add POST /goals/stop endpoint with AbortController cancellation
- Server: rewrite session middleware to direct DB token lookup
- Server: goals route fetches user's saved LLM config from DB
- Web: show installed apps in device detail Overview tab with search
- Web: add Stop button for running goals
- Web: replace API routes with remote commands (submitGoal, stopGoal)
- Web: add error display for goal submission failures
- Shared: add InstalledApp type and apps message to protocol
- Add device/session/step DB persistence in server agent loop
- Add goal preprocessor for compound goals (e.g., "open YouTube and search X")
- Add step-level logging to agent loop
- Fix dashboard WebSocket auth (direct DB token lookup instead of auth.api)
- Fix web layout to use locals.session.token instead of cookie
- Add dashboard-ws.svelte.ts WebSocket store with auto-reconnect
- Rewrite devices page with direct DB queries and real-time updates
- Add device detail page with live step display and session history
- Add Android companion app resources, themes, and screen capture consent
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Add deleteKeySchema for proper FormData validation
- Return { deleted: true } from deleteKey for change tracking
- Use $state for keysPromise to refresh list after create/delete
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Deploying to Railway instead of Cloudflare, so swap
@sveltejs/adapter-cloudflare for @sveltejs/adapter-node.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Server-side agent loop that adapts the CLI kernel to work over WebSocket.
Three new modules: stuck detection, LLM provider abstraction (OpenAI/Groq/
OpenRouter), and the main perception-reasoning-action loop. Also wires up
the goals route to start agent loops with duplicate-device protection.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
find_and_tap now scrolls down and rescans (up to 10 times) when the
target element isn't visible on the current screen. Stops as soon as
the element is found — no wasted scrolls. This removes the need for
LLMs to manually scroll-and-check in workflow prompts.
Also simplifies the Gemini-to-WhatsApp workflow prompts.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Multi-app workflow: ask Gemini about droidclaw.ai, copy response,
switch to WhatsApp, find contact, paste and send.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Groq (free tier, API key) is the primary setup path.
Ollama (local, no key) shown as secondary option.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Site and README now show both curl one-liner and manual steps.
Explicit note that bun is required (node/npm won't work).
Install script improvements: version check, node detection warning,
curl check, Windows detection, .git dir validation, shallow clone,
fallback on bun install --silent, post-install missing dep summary.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
One-liner install: curl -fsSL https://droidclaw.ai/install.sh | sh
Auto-installs bun and adb, clones repo, sets up .env.
Updated site meta tags and canonical URL to droidclaw.ai.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Reuses OpenAIProvider via Ollama's OpenAI-compatible API at localhost:11434.
No API key needed - just install Ollama, pull a model, and set LLM_PROVIDER=ollama.
Vision models (llava, llama3.2-vision) supported for screenshot fallback.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Update all GitHub links from thisuxhq to unitedbyai
- Add contributors section with sanju sivalingam and somasundaram mahesh
- Update meta author and footer attribution to unitedby.ai
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
New --flow mode executes scripted YAML steps without LLM, mapping 17
commands (tap, type, swipe, scroll, etc.) to existing actions. Element
finding uses accessibility tree text/hint/id matching.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- New actions: open_url, switch_app, notifications, pull_file, push_file, keyevent, open_settings
- Workflow system: runWorkflow() for multi-app sub-goal sequences with --workflow CLI flag
- Export runAgent() with {success, stepsUsed} return for workflow integration
- Fix clipboard_set shell escaping (single-quote wrapping matching skills.ts)
- Improve type action escaping for backticks, $, !, ?, brackets, braces
- Move parseJsonResponse to llm-providers.ts and export it
- Update SYSTEM_PROMPT and Zod schema for 22 total actions
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Removes the unnecessary nesting — all source, config, and docs now live
at the project root for simpler paths and commands.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Auto-detect screen resolution and compute dynamic swipe coordinates
- Detect foreground app each step via dumpsys activity
- Smart element filtering: deduplicate by position, score by relevance, compact to essentials
- Session logging with crash-safe .partial.json writes and final summary
- Real multimodal vision: send base64 screenshots to LLMs (off/fallback/always modes)
- Multi-turn conversation memory: maintain full chat history across steps with trimming
- Multi-step planning: think/plan/planProgress fields on every LLM decision
- Streaming responses for all 4 providers (OpenAI, Groq, OpenRouter, Bedrock)
- Comprehensive README with examples, architecture docs, and troubleshooting
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
