Vector Databases | Techlist.io

githubMar 19, 20263 min readCurated summary

How Squad runs coordinated AI agents inside your repository

Squad is an open-source GitHub Copilot project that places a preconfigured team of AI agents directly inside a repository. Rather than relying on a single chatbot or complex orchestration infrastructure, it coordinates specialized agents for design, implementation, testing, documentation, and review. Its core argument is that repository-native, versioned context makes multi-agent development more accessible, inspectable, and resilient. ## Coordinating Specialized Agents - Install Squad with `npm install -g @bradygaster/squad-cli`, then run `squad init` in a repository. - The setup creates roles such as lead, frontend developer, backend developer, tester, and documentation specialist. - A coordinator interprets natural-language requests, loads repository context, and assigns work to specialists. - Agents can work in parallel, create files and branches, write tests, and open pull requests. - They use shared decisions and project history rather than requiring every detail to be repeated in prompts. - Testing and review happen within the workflow: - Testers evaluate implementations and reject failing code. - A rejected author is prevented from revising its own work. - Another agent must address the problems, providing a more independent review. - Developers still answer questions, correct assumptions, and review and merge pull requests; Squad is collaborative orchestration rather than full autonomy. ## Repository-Based Shared Memory - Squad uses a “drop-box” model instead of depending on live chat synchronization or complex vector databases. - Architectural decisions, library choices, and conventions are appended to a versioned `decisions.md` file. - This creates: - Persistent shared knowledge - An understandable audit trail - Recovery after disconnects or restarts - Memory that can be reviewed and changed like code ## Replicating Context Across Agents - The coordinator remains a thin router instead of attempting to manage all implementation work. - Each specialist runs in its own inference call with an independent context window. - This replicates relevant repository context across agents rather than splitting one limited context among multiple roles. - Parallel, independent contexts reduce the risk that project-management instructions and other agents’ reasoning crowd out the actual coding task. - Supported models may provide context windows of up to 200,000 tokens. ## Versioned Agent Identities and History - Each agent’s behavior is primarily defined by repository files: - A charter describing its role and responsibilities - A history recording previous work - Shared team decisions - These files live in `.squad/` alongside the application code. - Cloning a repository also restores the team’s accumulated knowledge, making the agents effectively pre-onboarded. - Keeping memory in plain text makes it inspectable, versioned, and independent of hidden model state. ## Lowering the Barrier to Multi-Agent Development Squad’s main goal is to make agentic workflows practical without requiring users to build orchestration layers, configure databases, or master advanced prompt engineering. Its repository-native design favors simple setup, transparent memory, independent review, and recoverable project context. Developers interested in this approach can install Squad and experiment with it directly in the project repository.

Read original

Vector Databases AI Agents Github Copilot Prompt Engineering+4

dropboxJan 28, 20263 min readCurated summary

Engineering VP Josh Clemm on how we use knowledge graphs, MCP, and DSPy in Dash

Dropbox VP Josh Clemm argues that useful workplace AI requires a unified context engine capable of securely understanding and retrieving information across many SaaS applications. Dropbox Dash combines custom connectors, multimodal content processing, knowledge graphs, hybrid search, and personalized access-control-aware ranking to make that possible. Clemm favors indexed retrieval over purely federated approaches because preprocessing enables richer context, faster search, and company-wide access, though it requires substantial engineering and infrastructure. ## Building Dash’s Context Engine - Custom connectors crawl third-party applications while handling: - Rate limits - API differences - Application-specific permissions and ACLs - Incoming content is normalized, often into Markdown, and enriched with: - Titles and metadata - Extracted links - Embeddings - Other key information - Different media types require different processing: - Documents can be text-extracted and indexed. - Images may require CLIP or multimodal models. - PDFs combine text, figures, and other elements. - Audio is transcribed. - Videos may require scene-by-scene multimodal analysis when dialogue is insufficient. - Dash models relationships between content as a knowledge graph: - Meetings can connect to documents, participants, transcripts, and previous notes. - Cross-application relationships provide richer context for search and agents. - Data is stored in secure systems using: - BM25 lexical search - Dense-vector storage - Hybrid retrieval - Multiple ranking stages personalize results and enforce user-specific permissions. ## Indexed Retrieval Versus Federated Retrieval - Federated retrieval queries external systems at runtime. - Its advantages include: - Fast initial implementation - Minimal storage requirements - Relatively fresh data - Easy addition of MCP servers and connectors - Its drawbacks include: - Inconsistent API speed, quality, and ranking - Limited access to company-wide content - Expensive post-processing and reranking - Large token usage when agents reason over returned results - Indexed retrieval preprocesses content during ingestion. - Its advantages include: - Access to shared company connectors - Enriched datasets created offline - Faster queries - More opportunities for recall and ranking experiments - Its costs include: - Significant custom connector development - Freshness challenges - High hosting costs - Difficult storage and architecture choices involving vector search, BM25, hybrid retrieval, or graph RAG. ## Making MCP Practical at Scale - MCP can simplify tool integration, but tool definitions consume substantial context-window space. - Large tool descriptions and retrieval results contribute to context rot and reduce agent effectiveness. - Dash aims to limit context usage to roughly 100,000 tokens. - MCP-based agents can also be slow: simple queries may take up to 45 seconds, while direct index retrieval returns results within seconds. - Dropbox’s approach is to wrap its index in a consolidated “super tool,” reducing the need to expose many separate tools to the agent. ## Broader AI Engineering Practices - The talk also covers Dropbox’s use of: - LLMs as evaluators or judges - Prompt optimization with DSPy - Tool calling and MCP design - These techniques complement the underlying context engine rather than replacing the indexing, enrichment, graph modeling, and permission systems required for reliable workplace AI. A practical takeaway is that organizations building AI over proprietary data should treat retrieval as a full data-platform problem. Start with robust connectors and permissions, enrich content before retrieval, model relationships across sources, and use MCP selectively where indexed retrieval can provide faster and more controlled results.

Read original

Vector Databases Model Context Protocol RAG Multimodal AI+4