> ## Documentation Index
> Fetch the complete documentation index at: https://docs.formal.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# AI Satellite

> Release notes for Formal AI Satellite

<Update label="June 5, 2026" tags={["Improvements"]}>
  ## 1.4.0

  ### Changed

  * Treat default column label classifications as custom user-owned labels so they can be edited or deleted
</Update>

<Update label="March 10, 2026" tags={["Improvements"]}>
  ## 1.3.2

  ### Changed

  * Improve PHI classification for large individual JSON payloads.
</Update>

<Update label="March 10, 2026" tags={["Bug Fixes"]}>
  ## 1.3.1

  ### Fixed

  * Upgrade dependencies causing startup errors
</Update>

<Update label="March 9, 2026" tags={["New Features"]}>
  ## 1.3.0

  ### New

  * Upgrade PHI LLM model to a newer version
</Update>

<Update label="February 23, 2026" tags={["Bug Fixes"]}>
  ## 1.2.1

  ### Fixed

  * Fix PHI max output token cap limit errors
</Update>

<Update label="February 20, 2026" tags={["New Features"]}>
  ## 1.2.0

  ### New

  * General availability of stream session analysis for SSH, SSM, and Kubernetes exec sessions
</Update>

<Update label="February 12, 2026" tags={["Bug Fixes"]}>
  ## 1.1.1

  ### Fixed

  * Increased the default PHI max output token cap to 32768 to reduce truncation
</Update>

<Update label="January 10, 2026" tags={["New Features"]}>
  ## 1.1.0

  ### New

  * Enable classifications of database columns in conjunction with the data discovery satellite
</Update>

<Update label="November 7, 2025" tags={["Improvements"]}>
  ## 1.0.5

  ### Improved

  * Enhanced stability and performance
  * Optimized default configuration for production workloads
  * Simplified deployment with automatic GPU detection and configuration
</Update>

<Update label="November 7, 2025" tags={["Maintenance"]}>
  ## 1.0.4

  * Minor deployment improvements
</Update>

<Update label="November 3, 2025" tags={["New Features","Improvements"]}>
  ## 1.0.3

  ### New

  * Updated model to newer version
  * vLLM v0.11.0 upgrade
  * vLLM Engine 1 Usage
  * FP8 cache quantization for performance
  * GPU Parallelization Support
  * File processing support up to 45KB at a time

  ### Changed

  * Improved stability
  * Cache Stabilization
  * Prompt shortening + tweaks
  * Post-processing enhancements
</Update>

<Update label="October 14, 2025" tags={["Bug Fixes","Improvements"]}>
  ## 1.0.2

  ### Fixed

  * Token length handling fix
  * Label recognition improvements
</Update>

<Update label="October 7, 2025" tags={["New Features","Improvements"]}>
  ## 1.0.1

  ### New

  * Upgraded to 14B Instruct + Different Model Class

  ### Changed

  * Simplified PHI fields
  * Removed label normalization
</Update>

<Update label="September 25, 2025" tags={["New Features"]}>
  ## 1.0.0

  ### New

  * 12B parameter model
  * Full PHI/PII detection capabilities
  * Connector integration
</Update>