Urgent.News

650+ sources. One page. See who else covered it.

Editions

AI

ContextMemory v0.1.0-beta: what shipped

ContextMemory v0.1.0-beta is out. Agent memory you can open like a wiki — not a vector black box, not classic RAG inject. Release notes ## 0.1.0-beta (2026-08-15) Features add Cursor-style HTTP, vision, browser, PDF, and canvas agent tools. ( 0d322c2 ) derive LLM protocol capabilities from prompt profiles. ( bd159ad ) expand agentic guardrail catalog with enable/disable kinds ( 4fabaec ) expose…

ContextMemory v0.1.0-beta, a new release of the ContextMemory platform, was made available on August 15, 2026. This version introduces Cursor-style HTTP, vision, browser, PDF, and canvas agent tools, replacing the previous vector black box and classic RAG inject approaches. The release also includes enhancements to the LLM protocol capabilities derived from prompt profiles, expanded agentic guardrail catalog with enable/disable kinds, and the exposure of LLM generation defaults in Admin to apply to chat/generate payloads.

In addition, this version brings a multi-model harness, a live-data guardrail, and an Ollama num_ctx fix. Several bug fixes were also implemented, including stopping GITHUB_TOKEN from overriding PEER_SYNC_TOKEN on mirror push, coercing query_objects filter object/string into string array, and mapping fieldsToReturn alias in query_objects normalizer.

The release also includes the pinning of LinkedIn API version and clarification of dev.to errors, promotion of prose MCP tool JSON into structured tool_calls, and the regeneration of AddSkillActivation migration with model snapshot. Other improvements include rewriting SQL-style query_objects filters to Zuora field.OP:value, setting User-Agent on dev.to announcer requests, stabilizing Qwen client-side tools, and stopping capping Weak MCP catalogs.

Furthermore, this release stops live-data validation loops and softens noisy guardrail defaults, prevents RequireZeroExitCode guardrail from looping forever on retried tool failures, and stops tool_describe failures from poisoning agentic validation loops. Lastly, the LinkedIn release posts were tightened to better match the messaging voice.

Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at dev.to →

More in AI

More from Saturday 15 August →