Release: datasette 1.0a30 The big new feature in this alpha is a new customizable "Jump to..." menu, described in detail in The extensible "Jump to" menu in Datasette 1.0a30 on the Datasette blog. You can try it out by hitting / on latest.datasette.io - it looks like this: The ne
Release: datasette-agent 0.1a4 Taking advantage of the new makeJumpSections() JavaScript plugin hook added in Datasette 1.0a30 , datasette-agent now presents this "Start a new agent chat" interface as part of the Jump to menu, any time you hit / : You can try this out by signing
Release: datasette-fixtures 0.1a0 One of the smaller features in Datasette 1.0a30 is this: New documented datasette.fixtures.populate_fixture_database(conn) helper for creating the fixture database tables used by Datasette's own tests, intended for plugin test suites. This new pl
datasette-agent-charts 0.1a2 has been released, introducing "View SQL query" buttons beneath rendered charts. This update enhances the user experience by providing direct access to the underlying SQL queries that generate the visualizations. The release is tagged under datasette and datasette-agent.
→ This is a useful quality-of-life improvement for Datasette users, making it easier to inspect and understand the data behind charts, which could be valuable for debugging or learning in data exploration workflows.
Datasette Agent 0.1a3 has been released, featuring "View SQL query" buttons for both visible tables and collapsed SQL result tool calls. This update also improves handling of truncated responses, ensuring tables display even if SQL results are cut off, and refines the display of reasoning chunks. Datasette Agent is an extensible AI assistant designed for Datasette.
→ This update to Datasette Agent, an AI assistant for Datasette, is a high-signal release for Mark, as it enhances tooling for data exploration and RAG, directly supporting his interest in practical AI applications.
datasette-agent-charts 0.1a1 has been released, introducing enhanced chart visualization with magnitude-based shading for bar and waffle charts, and categorical coloring for text-based columns. The update also adds interactive tooltips and fixes a bug with waffleY chart descriptions. Security is improved by checking execute-sql permissions before querying column names.
→ This update to datasette-agent-charts improves data visualization and interaction within Datasette, which is highly relevant for anyone doing local data analysis or preparing datasets for LLMs like Gemma.
Hugging Face introduces Nemotron-Labs Diffusion Language Models, a new approach to text generation that leverages diffusion models, traditionally used for image generation. This method aims to significantly accelerate text generation speed by generating tokens in parallel rather than sequentially. Early results suggest a substantial speedup compared to traditional autoregressive models, potentially enabling near real-time text generation.
→ This could be a game-changer for local LLM inference, potentially making on-device models like Gemma and Llama vastly faster for real-time applications.
The article argues that specialized, smaller AI models often outperform larger, general-purpose models for specific tasks, offering better efficiency and cost-effectiveness. It highlights that procurement decisions frequently overlook this advantage, leading to over-reliance on large, expensive models when a fine-tuned, smaller alternative would suffice. The authors advocate for a strategic shift towards evaluating and deploying specialized AI solutions tailored to particular use cases.
→ This is a core tenet for local LLMs and on-device AI: smaller, specialized models like Gemma or Phi are often superior for specific tasks, aligning perfectly with the push for efficient, accessible AI.
tp3_memories_localgemma3:4b)