Scrapers that break on JS-heavy pages, get blocked, or lose progress when a run crashes? Crawlee for Python from @apify gives you one API for HTTP and headless browser crawling. An adaptive crawler decides when JS rendering is needed, with proxy rotation, retries, and a pausable request queue. It's open source and works with Playwright, BeautifulSoup, Parsel and Selenium.

Running a separate GPU for every fine-tuned LLM gets expensive fast. LoRAX from @predibase is a multi-LoRA inference server that serves thousands of fine-tuned models on a single GPU. It loads adapters on demand from HuggingFace or local disk and batches requests across them. Open source under Apache 2.0, with an OpenAI-compatible API, Docker images and Helm charts.

Turning a 200-page CIM into a screening memo by hand eats analyst hours. V7 Go from @v7labs automates document-heavy work in private markets, insurance and real estate with configurable AI agents (280+ templates) for IC memos, DDQs and LPA analysis. It's paid, with pricing via sales, plus audit logs, access controls and an MCP server. Connects to SharePoint, Box, Snowflake, PitchBook and Affinity.

Automations break when the AI doesn't know your company's docs, and IT won't approve tools without access controls. Cassidy from @cassidy_ai is a no-code agent and workflow platform that grounds agents in a synced knowledge base, with human-in-the-loop approvals. It's freemium, includes SSO, RBAC and audit logs, and connects to Slack, Salesforce, HubSpot, SharePoint, Okta and Azure AD.

New on EveryDev.ai: Yellow.ai Enterprise agentic AI platform for building and deploying AI agents that automate customer and employee service across.

Need GPUs for inference or training but want your data to stay with a European provider? @Scaleway offers GPU Instances (L4, L40S, H100, B300-SXM), serverless Generative APIs and dedicated inference deployments, alongside managed Kubernetes, databases and bare metal. It's paid, managed through API, CLI or Terraform, and includes an Amazon S3-compatible object storage API.

Generating embeddings shouldn't mean installing a heavy PyTorch stack just to index some documents. FastEmbed is a lightweight Python library that runs on ONNX Runtime with quantized weights. It covers dense, sparse (SPLADE++), ColBERT, image and ColPali embeddings plus cross-encoder rerankers. It's open source, with GPU support via fastembed-gpu and integrations for Qdrant, LangChain and LlamaIndex.

Losing action items because nobody took good notes in the meeting? Notta from @NottaOfficial records and transcribes calls, with a bot or bot-free on desktop, then writes summaries and action items you can search later. It's freemousing with a free plan, and it connects to Zoom, Google Meet, Teams, Slack, Notion, Salesforce, HubSpot, Zapier and offers MCP & CLI for AI assistants.

Your docs, PDFs and slides pile up, and RAG over a vector database still misses the connections between them. OpenKB from @PageIndexAI is a CLI that compiles raw documents into an interlinked Markdown wiki with summaries, concept pages and entity pages, using PageIndex tree retrieval instead of vectors. It works with OpenAI, Claude and Gemini via LiteLLM, and the output opens in Obsidian. Open source under Apache 2.0.

Waiting on GPU quota or stitching together separate vendors for training, inference and storage? Verda from @verdacloud is a full-stack AI cloud with on-demand GPU and CPU instances, serverless auto-scaling GPU containers, NVMe block storage, a POSIX shared filesystem and an OCI registry. It's paid, with a self-service API, and instant clusters come with InfiniBand.