Skip to content
ZYNXTECH

Log / 15 entries

Release log

Versions and publications across every project, newest first. Version notes come from each project's changelog, which has the full detail.

November 20268 entries

  1. LLMDynov0.6.1On npm, published with provenance

    • The release workflow publishes the npm package through trusted publishing, so no token is stored.
    • Fixes a JavaScript client call that threw instead of returning a rejected promise.
    Source
  2. LLMDynov0.6.0LLMDyno for JavaScript and TypeScript

    • npx llmdyno starts the server; npm install llmdyno gives a typed client for Node 22 and later.
    • All 116 web API operations are typed under dyno.api, generated from the server routes.
    • Benchmark your own TypeScript agent (Vercel AI SDK, Mastra, LangGraph.js, OpenAI Agents SDK) in the sandbox.
    Source
  3. pbidriverv0.1.1On npm, plus a documentation site

    • npx pbidriver runs the same CLI from Node.js through uv or Python, with a JavaScript run() API.
    • Docs at speak2jerome.github.io/pbidriver, with setup for Claude Code, GitHub Copilot, Codex and Cursor.
    • pywin32 is only required on Windows, so the package installs in CI and docs builds.
    Source
  4. pbidriverv0.1.0First public release

    • Captures of the canvas, single visuals and every view, including Table, Model, DAX query and TMDL.
    • Reads chart points, cards, tables, matrices and slicers. Answers all three kinds of Power BI dialog.
    • Developed under the working name pbi_shot.
    Source
  5. LLMDynov0.5.0The library can do everything the dashboard does

    • All 116 web API operations have a Python call and a CLI command, kept in step by tests.
    • Integrations for pandas, polars, MLflow, Weights & Biases, Langfuse, OpenTelemetry, LiteLLM, Inspect AI and pytest.
    • Remote mode: run the server on a GPU box and drive it from another machine with a token.
    Source
  6. LLMDynov0.3.1On PyPI

    • pip install llmdyno. Releases publish through trusted publishing.
    Source
  7. LLMDynov0.3.0Python library

    • import llmdyno: cloud_model(), local_model(), bench() with a preflight, and Results with to_pandas().
    Source
  8. LLMDynov0.2.0Apple Silicon, ARM64, SWE-bench and Aider polyglot

    • Runs on macOS, Linux ARM64 and Windows on ARM. Plans memory around the Metal working-set limit.
    • Imports SWE-bench (Verified, Lite, Verified mini) graded by the official harness, and the 225 polyglot exercises.
    • Shows Artificial Analysis reference numbers next to your own measurements.
    Source

October 20261 entry

  1. LLMDynov0.1.0First public release

    • VRAM projection from the GGUF header, seven agent harnesses in offline sandboxes, blind LLM judging and a results dashboard.
    • Ships with sample results: 15 configurations on an RTX 5080.
    Source

July 20266 entries

  1. FinventorySource published on GitHub

    • Django 5.1 and React 19 personal finance platform for the Indian market.
    Source
  2. PerfReviewSource published on GitHub

    • Flask and React appraisal platform with self, manager and HR review stages.
    Source
  3. MindWorkSource published on GitHub

    • Django mental health toolkit with validated screenings and a rule-based insight engine.
    Source
  4. Wealth Management 360Report published on GitHub

    • Six-page Power BI report for a wealth management business.
    Source
  5. HR AnalyticsReport published on GitHub

    • Eight-page attrition report with key influencers, a Python correlation heatmap and Q&A.
    Source
  6. HR RecruitmentReport published on GitHub

    • Seven-page candidate pipeline report covering skills, salary, location and diversity.
    Source