{"id":48878821,"url":"https://github.com/pmbstyle/fara-agent","last_synced_at":"2026-04-16T01:34:47.552Z","repository":{"id":326151666,"uuid":"1104220555","full_name":"pmbstyle/fara-agent","owner":"pmbstyle","description":"A local browser automation agent based on Microsoft Fara-7B model optimized for LM Studio inference.","archived":false,"fork":false,"pushed_at":"2025-11-25T23:43:36.000Z","size":19,"stargazers_count":6,"open_issues_count":1,"forks_count":2,"subscribers_count":0,"default_branch":"main","last_synced_at":"2025-11-29T03:43:20.438Z","etag":null,"topics":["ai-agents","browser-automation","browser-use","computer-use","fara","lm-studio","playwright","vision-agents"],"latest_commit_sha":null,"homepage":"","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/pmbstyle.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2025-11-25T23:21:05.000Z","updated_at":"2025-11-28T20:42:40.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/pmbstyle/fara-agent","commit_stats":null,"previous_names":["pmbstyle/fara-agent"],"tags_count":null,"template":false,"template_full_name":null,"purl":"pkg:github/pmbstyle/fara-agent","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/pmbstyle%2Ffara-agent","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/pmbstyle%2Ffara-agent/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/pmbstyle%2Ffara-agent/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/pmbstyle%2Ffara-agent/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/pmbstyle","download_url":"https://codeload.github.com/pmbstyle/fara-agent/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/pmbstyle%2Ffara-agent/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":31867712,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-04-15T15:24:51.572Z","status":"ssl_error","status_checked_at":"2026-04-15T15:24:39.138Z","response_time":63,"last_error":"SSL_read: unexpected eof while reading","robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":false,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["ai-agents","browser-automation","browser-use","computer-use","fara","lm-studio","playwright","vision-agents"],"created_at":"2026-04-16T01:34:47.366Z","updated_at":"2026-04-16T01:34:47.530Z","avatar_url":"https://github.com/pmbstyle.png","language":"Python","readme":"# Fara Browser Automation Agent\n\nA local browser automation agent based on [Microsoft Fara-7B](https://www.microsoft.com/en-us/research/blog/fara-7b-an-efficient-agentic-model-for-computer-use/) model optimized for [LM Studio](https://lmstudio.ai/) inference.\n\nRun browser automation locally on a consumer-grade GPU with a variation of quantized models.\n\n## Features\n\n- ✅ 100% local AI browser agent\n- ✅ Quantized models support\n- ✅ Completely self-contained (no external dependencies)\n- ✅ Optimized for LM Studio\n- ✅ Browser automation via Playwright\n\n\u003ca href=\"https://www.youtube.com/watch?v=cM-ApEK2biw\"\u003e\n   \u003cimg width=\"846\" height=\"476\" alt=\"image\" src=\"https://github.com/user-attachments/assets/ef55585b-3904-439f-badb-0f39e88fbbd1\" /\u003e\n\u003c/a\u003e\n\n\n## Setup\n\n### 1. Install Dependencies\n\n```bash\npip install -r requirements.txt\nplaywright install firefox\n```\n\n### 2. Setup LM Studio\n\n1. Download and install [LM Studio](https://lmstudio.ai/)\n2. Download the [Fara-7B](https://huggingface.co/bartowski/microsoft_Fara-7B-GGUF) model (GGUF format):\n   - Search for: `microsoft_fara-7b`\n   - Recommended: Q5_K_M quantization (6GB)\n3. Load the model in LM Studio\n4. Start the local server (default port: 1234)\n5. In model settings:\n   - Context Length: 8192+\n   - Temperature: 0.0\n   - Top P: 0.9\n\n### 3. Run the Agent\n\n```bash\npython run_agent.py --task \"Go to wikipedia.org and search for cats\" --headful\n```\n\nOptional debug flags (enabled by default in headful mode):\n- `headful`: displays a browser window\n- `show_overlay`: bottom-right HUD with latest model responses (hidden during screenshots)\n- `show_click_markers`: transient markers for clicks/hover/type coordinates (hidden during screenshots)\n\n## Configuration\n\nEdit `config.json` to change:\n- Model endpoint (default: http://localhost:1234/v1)\n- Model name\n- Max rounds\n- Screenshot settings\n- Max images to keep in context (`max_n_images`, default 1)\n- Downloads folder for saving files\n- Debug overlay and click markers (`show_overlay`, `show_click_markers`)\n\n## How It Works\n\n1. **Browser Control**: Uses Playwright to control Firefox\n2. **Vision**: Takes screenshots and sends them to the model\n3. **Actions**: Model returns tool calls (click, type, scroll, navigate, hover, keypress, wait, memorize facts)\n4. **Single-Image Mode**: Only sends the latest screenshot to LM Studio (better compatibility)\n5. **Loop Guard**: Tracks scroll position and warns the model when it oscillates up/down\n\n## Limitations\n\n- Quantized models have reduced capabilities vs full model\n- LM Studio has issues with multiple images in the conversation history\n- Some complex tasks may cause loops (scrolling, navigation)\n\n## Troubleshooting\n\n**Browser not visible?**\n- Make sure you're using `--headful` flag\n\n**Model not responding?**\n- Check LM Studio server is running on port 1234\n- Verify model is loaded in LM Studio\n\n**Agent looping?**\n- Try reducing the temperature in LM Studio to 0.0\n- Reduce `max_rounds` in config.json\n\n## License\n\nMIT License - Based on [Microsoft Fara-7B](https://www.microsoft.com/en-us/research/blog/fara-7b-an-efficient-agentic-model-for-computer-use/)\n\n","funding_links":[],"categories":[],"sub_categories":[],"project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fpmbstyle%2Ffara-agent","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fpmbstyle%2Ffara-agent","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fpmbstyle%2Ffara-agent/lists"}