π ASSISTANT
Editorial Copilot & Post Creator β Cloud & Nano AI
Context-aware reply drafting and viral post creation directly inside π (Twitter) with zero API paywalls. Powered by official Google Gemini Cloud REST API (gemini-3.1-flash-lite, <220ms), on-device Gemini Nano, on-screen Floating HUD, and mandatory human review before publishing.
Amplifying Human Voice, Not Replacing It
Most social AI tools attempt to automate speech: churning out robotic, sycophantic replies that pollute public discourse. π Assistant is strictly an editorial copilot.
Reads author, text, thread context, and intent directly from π DOM without leaving the feed.
User selects desired stance: Insightful, Supportive, Witty, Counter-point, or Question.
Gemini Cloud REST API (<220ms) or local Gemini Nano synthesizes reply with intelligent cascade fallback.
User reviews, edits, refines, and approves the text. The assistant never posts on your behalf.
Approved text drops into the native π composer. The user retains sole publication authority.
Sub-Second Interaction
By utilizing on-device Gemini Nano when available, drafting latency drops below 250 milliseconds. The interaction feels like an extension of thought rather than a round-trip network query.
Zero $100/mo API Paywall
Operates entirely via client-side DOM injection and your existing authenticated sessions. No Twitter/X Enterprise API subscriptions, no monthly tokens, and no commercial credit cards.
MIT Open-Source Code
Every single line of extension code is publicly auditable on GitHub. You can inspect the source code, verify that zero telemetry scripts exist, or build the extension from source.
Experience the Copilot Causality Flow
Simulate how π Assistant analyzes context, drafts tone-calibrated responses across four modular AI engines, and presents them for human approval.
Four Architectural Pathways & Intelligent Cascade Fallback
π Assistant implements four distinct AI execution pathways with an automated cascade failover system. From cloud-speed REST inference to local silicon and browser automation, every environment is supported with zero middleman servers.
01 / Cloud REST API
True asynchronous REST API integration calling Google's official generativelanguage.googleapis.com endpoint directly.
- Zero tab overhead: Pure background HTTP fetch without opening any browser tabs.
- Lightning latency: Under 220ms response time powered by
gemini-3.1-flash-lite. - BYOK security: Uses free Google AI Studio key stored strictly in
chrome.storage.local. - Multi-model choice: Select between Flash Lite, 2.5, 2.0, or 1.5 Pro on the fly.
02 / Gemini Nano
Runs directly on the user's physical device via Chrome's experimental window.ai (Prompt API).
- Zero external network calls: Inference executed entirely on local silicon.
- Sub-second response: 150msβ250ms average generation time.
- Offline capability: Operates without an active internet connection.
- Hardware dependency: Requires Chrome 128+ with Prompt API enabled.
03 / Headless Direct
Communicates directly with Google Gemini using the user's active browser Google session cookies.
- No third-party proxy: Encrypted client-to-Google session handshake.
- Zero API subscription: Bypasses Twitter's $100/mo API paywall completely.
- Zero tab overhead: Operates silently in background worker without tabs.
- Frontier reasoning: Direct Gemini 1.5 / 2.0 server-side intelligence.
04 / Web Tab Automation
Scripted browser automation interacting with the web interface at gemini.google.com.
- Universal fallback: Ensures generation for any user without API key or Nano hardware.
- Zero setup required: Operates with existing web login session.
- Temporary tab lifecycle: Opens a background tab, extracts response, and auto-closes.
- Browser dependency: Requires active login on gemini.google.com.
Engine Comparison Matrix
| Attribute | Cloud REST API (Primary) | Gemini Nano (On-Device) | Headless Direct | Web Tab Automation |
|---|---|---|---|---|
| Execution Location | Google Cloud REST API | Local Device Silicon (GPU / NPU) | Google Cloud Infrastructure | Google Web Interface |
| Network Transmission | REST to generativelanguage.googleapis.com | 0 Requests (100% Offline) | Direct session endpoints | Browser Tab to gemini.google.com |
| Average Latency | <220ms (Fastest Cloud) | <250ms (Instantaneous) | ~650ms β 850ms | ~1,200ms β 1,800ms |
| Browser Tab Overhead | 0 Tabs (Direct async fetch) | 0 Tabs (Local window.ai) | 0 Tabs (Service worker) | 1 Auto-Closing Tab |
| Authentication / Keys | Free Google AI Studio API Key | None (Chrome 128+ window.ai) | Google Browser Session Cookie | gemini.google.com Web Login |
| API Subscription Cost | $0 / mo (Free Tier) | $0 / mo | $0 / mo | $0 / mo |
| Third-Party Intermediaries | None (Direct client-to-Google) | None (Strict client sandbox) | None (Direct client-to-Google) | None (Direct client-to-Google) |
On-Screen Floating HUD & Viral Creator Studio
v1.3.0 brings a persistent, non-intrusive HUD directly onto the π feed alongside an AI Post Creator powered by 7 battle-tested viral archetypes.
The Draggable Timeline HUD
Never leave the timeline to tweak generation parameters. The Floating HUD lives as a lightweight, glassmorphic dock directly in the π DOM. Drag it anywhere, collapse it to a discrete pill, or switch tones on the fly.
- Live Engine Telemetry: Monitor active engine, token latency, and failover status in real-time.
- 1-Click Tone Switcher: Instantly toggle between Quick, Thoughtful, Witty, or Custom personas.
- Custom Prompt Editor: Adjust temperature, post-processing filters, or persona instructions directly inside the timeline.
- Minimal Footprint: Zero layout shift on x.com, auto-collapsing when scrolling deep feeds.
7 Calibrated Viral Post Archetypes
Generate original long-form posts, threads, and high-signal insights engineered for high engagement and zero corporate fluff.
Hooks attention with an immediate high-value framework, bullet points, and step-by-step technical breakdown.
Challenges popular timeline dogmas with nuanced, defensible counter-arguments and firsthand builder lessons.
Transforms raw production failures, pivots, and architectural milestones into relatable narrative storytelling.
Curates high-utility developer tools, libraries, and workflow enhancements structured for bookmarking and saving.
Sparks deep community discussion with provocative, open-ended technical dilemmas that invite builder replies.
Distills complex industry benchmarks, telemetry, and metrics into digestible takeaways with clear executive summaries.
Inspires builders, founders, and engineers to push beyond creative limitsβgrounded in real craft, discipline, and execution without superficial platitudes.
Six Context-Aware Reply Personas
Trained prompt engineering avoids generic AI clichΓ©s ("Great thread!", "Delve into"). Each persona is calibrated for authentic human discourse.
Concise, conversational, and direct. Ideal for fast mobile acknowledgments, casual camaraderie, and friendly feed interactions without flowery prose.
Adds intellectual value, cites systemic second-order effects, and synthesizes technical nuances. Perfect for establishing credible expertise on tech timelines.
Validates the author's premise while reinforcing it with a complementary example or shared builder experience. Constructive and encouraging.
Dry, tech-insider satire and clever wordplay. Lightens intense timeline debates without descending into cheap sarcasm or insults.
Respectful, rigorous intellectual pushback. Identifies edge cases, alternative architectures, or unexamined assumptions to foster genuine debate.
Poses a penetrating open-ended inquiry derived directly from the post's core claim. Designed to invite author replies and thread amplification.
All generated replies pass through client-side regex cleansing filters that automatically strip model thinking scratchpads
(<thought> blocks), conversational preambles ("Here is a reply:"), and extraneous asterisks. Replies are 100% natural, polished, and ready for immediate human review.
Transparent, Private, Verifiable
We believe browser extensions require the highest degree of security disclosure. π Assistant operates with complete cryptographic integrity.
No Middleman Server
GMD operates no central API relay for π Assistant. Requests go either straight to your local device NPU (Gemini Nano) or directly to Google. Your tweets are never routed through our hardware.
No Credential Storing
The extension never asks for, captures, or persists passwords, tokens, or financial credentials. Authentication relies entirely on your browser's existing secure cookie jar.
100% Open Source
Published on GitHub under the permissive MIT license. Anyone can audit the content scripts, verify
permissions in manifest.json, or contribute improvements.
Technical Specifications
| Target Platform | π (formerly Twitter) web interface (x.com / twitter.com) |
|---|---|
| Browser Compatibility | Google Chrome 128+ (for on-device Nano), Brave, Arc, Edge |
| Architecture | Manifest V3 (Content Script DOM Injector + Service Worker) |
| AI Inference Models | Google Gemini Cloud REST API (gemini-3.1-flash-lite) / Gemini Nano (Local) / Headless Direct / Web Tab Automation |
| Source Repository | https://github.com/gomedz/twitter-auto-reply |
| Software License | MIT Open Source License |
| Product Maturity | Production Release v1.3.0 (Verified & Published) |
Elevate Your Timeline Presence
Install π Assistant from the Chrome Web Store or inspect the complete open-source repository on GitHub.