Decision: prioritize fetch reliability over speed for agents that need consistent page retrieval across diverse domains.
Hardware support: supports Colab T4 for 13M model and A100/RTX4090 for billion‑parameter models.
Features: end‑to‑end pipeline, SFT & RLHF notebooks, tiktoken r50k_base tokenization.
In September 2026 I posted about seven open‑source AI agents and two related tools, highlighting a Voice‑AI Booking Agent, a Self‑Directed Data Analyst, the JEV Decision‑Only Judge, an AI Engineering Curriculum, the Jev Decision Model, the Understand‑Anything code knowledge‑graph, and Apify Web Fetch.
The Voice‑AI Booking Agent runs on GLM‑5.3 Flash through Telnyx, about $0.30 per booking call.
Its workflow is: caller → Voice Agent → Check availability → Book slot → Create calendar invite → Store CRM record → Generate follow‑up.
The Self‑Directed Data Analyst uses Liner Mark 1.0 to route each request to an appropriate underlying model while exposing a single model interface to the agent.
JEV came in at about $0.04 per 1,000 judgments with roughly 0.15 s median latency.
With a 0.9 confidence threshold, the pooled cascade escalated about 34% of the judgments and retained 99.6% of GPT‑6's accuracy at 47% of its fee.
The AI Engineering Curriculum contains 523 lessons across 20 phases.
Each lesson follows the pattern: Problem → Concept → Build It → Use It → Ship It.
Jev removes string generation from that loop.
Jev returns typed answers with probabilities that the application can use directly.
Understand‑Anything is a Claude Code plugin that builds an explorable map of your project.
Apify Web Fetch handles proxy rotation, browser fingerprinting, JS rendering, challenge solving, retries.