Venster

Turn code discussions into searchable engineering knowledge, directly inside VS Code.

Venster is a VS Code extension and web platform that lets development teams create, discuss, and resolve code-level issues — called Capsules — directly attached to source code. When a Capsule is resolved, an AI knowledge engine automatically extracts structured knowledge and stores it with vector embeddings for semantic search, turning every resolved issue into reusable team knowledge.

---

The Problem

Important debugging conversations happen in Slack, meetings, GitHub issues, or email threads — far away from the code they're about. The result:

• Context is lost. A developer discovers a tricky piece of code but the explanation lives in a chat message from months ago.
• Knowledge evaporates. Team members leave, onboarding docs go stale, and tribal knowledge disappears.
• Problems repeat. The same null pointer exception or race condition gets debugged by three different people on three different teams.
• Discussions are disconnected from code. A code review comment on line 42 can't capture the broader architectural discussion that led to that decision.

---

The Solution

Venster connects engineering conversations directly to code through Capsules — lightweight discussion threads anchored to specific files, line ranges, and code selections. Every resolved Capsule automatically becomes searchable knowledge, powered by vector embeddings and AI-generated summaries.

The lifecycle:

Select code → Create Capsule → Discuss with team → Resolve → Knowledge is captured → Searchable forever

---

Key Features

🔗 Code-Context Capsules

Each Capsule is anchored to a specific file, line range, and selected code snippet. Your team sees exactly which code the discussion is about — no copy-pasting snippets into chat.

How to use: Select code in the editor, right-click, and choose Venster: Create Capsule from Selection. Or use the command palette (Ctrl+Shift+P → "Venster").

💬 Team Discussions

Add comments to any Capsule. Teammates can reply with solutions, ask clarifying questions, and build on each other's insights — all within the context of the code.

📋 Capsule Status Workflow

Capsules follow a structured lifecycle that keeps discussions moving:

OPEN → IN_REVIEW → ANSWERED → RESOLVED
                    ↘           ↗
                    ARCHIVED (terminal)

• OPEN — New Capsule, waiting for discussion
• IN_REVIEW — Active discussion
• ANSWERED — Solution proposed, awaiting resolution
• RESOLVED — Final resolution captured
• ARCHIVED — Closed without resolution

🧠 Automatic Knowledge Extraction

When a Capsule is resolved, the AI knowledge engine automatically extracts structured knowledge:

• Title — What was the problem
• Summary — Brief description
• Root Cause — Why it happened
• Solution — How it was fixed
• Category — Bug, Performance, Security, Design, Configuration, etc.
• Tags — Searchable labels
• Confidence score — Quality indicator

🔍 Semantic Knowledge Search

Search your team's resolved issues using natural language. Venster uses pgvector embeddings for semantic similarity search — find answers by describing the problem, not just by keywords.

How to use: Run Venster: Search Knowledge from the command palette, or open the Knowledge Base in the Venster webview.

🤖 AI-Powered Answers

When you search, Venster generates an AI-drafted answer from your team's resolved issues, with inline citations. If your team hasn't resolved the issue yet, it falls back to relevant Stack Overflow threads.

🌐 Community Insights

Published knowledge items are visible to all Venster users across teams. Browse solved problems shared by the community, or search for specific solutions — no workspace required.

❓ Community Q&A

Ask and answer engineering questions in a public forum. Tag questions, accept answers, and build a shared knowledge base beyond your immediate team.

🔔 Notifications

Stay updated with a badge counter in the VS Code status bar. The extension polls for new notifications every 30 seconds when you're logged in. A full notification center is available in the Venster webview.

🏗️ Workspaces & Projects

Organize your team's Capsules into workspaces and projects. Manage team membership and control who can see and resolve Capsules.

🔁 Duplicate Detection

Before creating a new Capsule, Venster checks for similar resolved issues in your workspace using vector similarity — preventing duplicate discussions and surfacing existing solutions.

---

How Venster Works

Developer finds a tricky piece of code
        ↓
Selects the relevant code
        ↓
Creates a Capsule (with file, line range, and code snippet attached)
        ↓
Adds a question or context
        ↓
Teammates view the Capsule and discuss
        ↓
A reviewer is assigned from the project team
        ↓
Problem is resolved with a final solution
        ↓
AI extracts structured knowledge automatically
        ↓
Knowledge becomes searchable for the entire team (and community, if published)

---

Working With Your Team

Creating a Capsule

1. Select the code you have a question about
2. Right-click → Venster: Create Capsule from Selection
3. Enter a title, priority, and project
4. The Capsule appears in the Capsule tree view and the Venster webview

Discovering Capsules

• Tree View: The Venster sidebar shows all Capsules for your projects, sorted by newest first
• Per-File View: Run Venster: View Capsules for Current File to see Capsules attached to the file you're editing
• Dashboard: The Venster webview dashboard shows trending Capsules, aging Capsules, and hot files

Collaborating

1. Developer A creates a Capsule around a confusing piece of code
2. Developer B opens the Capsule, sees the exact code context, and replies with an explanation
3. Reviewer (assigned from the project team) validates the solution
4. The Capsule is marked RESOLVED with a final solution

Preserving Knowledge

Once resolved, the AI engine extracts a structured knowledge item with root cause, solution, tags, and confidence score. This knowledge is:

• Searchable by your team via semantic search
• Publishable to Community Insights for cross-team sharing
• Citable in future AI-drafted answers

Future Developer Benefit

A new team member encounters the same confusing code. They search Venster Knowledge: "Why does this authentication check exist?" — and find the exact resolution from the original discussion, with the code context preserved.

---

Example Use Case

Scenario: A developer encounters an unexpected null return from getUserPreferences() in src/services/UserService.java.

1. Select code: Highlight the method call
2. Create Capsule: "getUserPreferences() returns null unexpectedly"
3. Add context: "This causes a NullPointerException in the dashboard renderer on line 142"
4. Teammate responds: "This was added in the v2.3 migration — the method now returns null for new users instead of an empty object. You need to add a null check."
5. Resolve: The reviewer confirms the explanation, marks it RESOLVED
6. Knowledge captured: Root cause documented as "v2.3 migration changed return type behavior," solution documented as "Add null check before passing to renderer"
7. Next developer: Searches "getUserPreferences returns null" and finds the answer immediately

---

Getting Started

Installation

1. Open VS Code
2. Open the Extensions panel (Ctrl+Shift+X)
3. Search for "Venster"
4. Click Install
5. Open the Venster view in the Activity Bar (left sidebar)
6. Click Venster: Login from the command palette and enter your credentials
7. Start creating Capsules!

First Capsule

1. Open any source file
2. Select a few lines of code
3. Press Ctrl+Shift+P → Venster: Create Capsule from Selection
4. Enter a title and priority
5. Your Capsule is created and appears in the sidebar

---

Commands

All commands are accessible via the Command Palette (Ctrl+Shift+P):

Command	Description
Venster: Login	Sign in with your email and password
Venster: Logout	Sign out and clear stored credentials
Venster: Create Capsule from Selection	Create a new Capsule anchored to your current code selection
Venster: View Capsules for Current File	List all Capsules attached to the file you're editing
Venster: Reply to Capsule	Add a comment to an existing Capsule
Venster: Refresh Capsule List	Reload the Capsule tree view
Venster: Search Knowledge	Search your team's resolved issues using natural language
Venster: Open Venster UI	Focus the full Venster webview panel
Venster: Create Capsule in Venster UI	Open the Capsule creation form in the webview, pre-filled with your current editor context

Context Menu

When you have code selected in the editor, right-click to see Venster: Create Capsule from Selection in the context menu.

---

Capsule Workflow

1. Select Code

Highlight the relevant code in your editor. The extension captures the file path, line range, and selected text automatically.

2. Create a Capsule

Use the command palette or context menu. Provide:

• Title — What's the question or issue
• Priority — LOW, MEDIUM, or HIGH
• Project — The project this Capsule belongs to

3. Add Context

The Capsule includes your code snippet and file location. Add a description explaining the problem or question.

4. Team Discussion

Teammates view the Capsule, see the exact code context, and reply with explanations, suggestions, or follow-up questions.

5. Assign Reviewer

A reviewer from the project team is assigned to validate the solution.

6. Resolve

When the discussion reaches a conclusion, the reviewer (or a workspace admin) marks the Capsule as RESOLVED with a final solution summary.

7. Knowledge Captured

The AI engine automatically extracts structured knowledge from the resolution — title, summary, root cause, solution, category, and tags — and stores it with a vector embedding for future search.

---

Knowledge Search

How It Works

Venster stores each knowledge item with a 1536-dimensional vector embedding generated from the summary and solution text. When you search, your query is embedded and compared using cosine similarity via pgvector — finding results by meaning, not just keywords.

What's Searched

• Your workspaces — Resolved Capsules from your team's projects
• Community Insights — Public knowledge items shared across all teams
• Stack Overflow — Relevant public threads (as a fallback when your team hasn't resolved the issue)

AI-Generated Answers

The search results include an AI-drafted answer synthesized from your team's top-matching resolved issues, with inline citations. When the team has no match, it falls back to Stack Overflow references.

Knowledge Categories

• BUG
• PERFORMANCE
• SECURITY
• DESIGN
• CONFIGURATION
• TESTING
• DOCUMENTATION

Publishing Knowledge

Workspace admins and owners can publish resolved knowledge items to Community Insights, making them searchable by all Venster users. Source code and workspace details are redacted in public items — only the knowledge summary is shared.

---

Architecture

┌──────────────────┐
│    VS Code       │
│  ┌────────────┐  │
│  │  Extension  │──┼──→ Command Palette / Context Menu / Tree View
│  │  (TypeScript)│  │
│  └──────┬─────┘  │
│         │        │
│  ┌──────▼─────┐  │
│  │  React      │  │
│  │  Webview    │  │──→ Dashboard / Capsules / Knowledge / Community / Q&A
│  │  (Vite)     │  │
│  └──────┬─────┘  │
└─────────┼────────┘
          │ REST API
┌─────────▼────────┐
│  Spring Boot     │
│  Backend         │──→ Auth / Capsules / Knowledge / Notifications / Q&A
│  (Java 17)       │
└─────────┬────────┘
          │
┌─────────▼────────┐
│  PostgreSQL      │
│  + pgvector      │──→ Vector similarity search (1536-dim embeddings)
│  (v16)           │
└──────────────────┘

Components

Component	Technology	Description
VS Code Extension	TypeScript	Commands, tree view, editor integration, notifications
React Webview	React 19, TypeScript, Tailwind CSS	Full Venster UI embedded in VS Code sidebar
Backend API	Java 17, Spring Boot 3.2, Spring Security	REST API with JWT authentication
Database	PostgreSQL 16 + pgvector	Data storage with vector similarity search
AI Engine	LLM (Groq) + local embeddings	Knowledge extraction and semantic search

---

Configuration

The extension connects to the Venster backend service. You can configure the backend URL in VS Code settings:

{
  "microhubs.baseUrl": "https://venster-backend.onrender.com"
}

The default points to the hosted production backend. Override this for local development:

{
  "microhubs.baseUrl": "http://localhost:8082"
}

---

Troubleshooting

"Please login first" error

Run Venster: Login from the command palette and enter your credentials. Your session token is stored securely in VS Code's SecretStorage.

Capsule list is empty

Make sure you're logged in and have created a Capsule. The tree view refreshes automatically after login and Capsule creation.

Backend connection issues

If the extension can't reach the backend, check your internet connection. The default backend URL is https://venster-backend.onrender.com. If you're developing locally, update the microhubs.baseUrl setting to point to your local backend.

Knowledge search returns no results

Knowledge items are generated from resolved Capsules. If you haven't resolved any Capsules yet, the knowledge base will be empty. Try publishing knowledge to Community Insights to browse public items.

Webview not loading

Run Venster: Open Venster UI from the command palette to focus the Venster panel. If it still doesn't appear, try reloading the VS Code window (Developer: Reload Window).

---

Security

• Authentication tokens are stored in VS Code SecretStorage (encrypted, not in settings files)
• All API communication uses HTTPS (production)
• The webview uses a strict Content Security Policy (CSP)
• Public knowledge items are redacted — source code and workspace details are never shared publicly
• The extension only accesses files within your workspace

---

Feedback & Contributions

Report issues and provide feedback on GitHub:

Repository: github.com/Shivajikompelli/venster

---