Skip to content

v0.9.0

Latest
Latest

Choose a tag to compare

View all tags
@karnsab karnsab released this 18 Feb 13:03
· 1 commit to main since this release
v0.9.0
384782b
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Release Notes: v0.9.0

Release Date: February 2026
Previous Version: v0.8.1 (December 2, 2025)
Branch: main

Summary

Version 0.9.0 is a major platform update delivering significant architectural changes, new features, and infrastructure modernization. With 162 commits, 358 files changed, and 6 merged pull requests, this release represents the largest evolution of the Amplify backend since launch.

Highlights at a Glance

  • ⚠️ Breaking Change — All shared environment variables migrated to AWS Parameter Store
  • ⚠️ Basic-Ops Elimination — amplify-lambda-basic-ops service removed; existing deployments must migrate
  • 🚀 21 New AI Models — GPT-5 series, Claude 4.x, Gemini 2.x, Amazon Nova, Mistral Pixtral
  • ⚡ API Gateway Streaming — Migrated from Function URLs to API Gateway Lambda response streaming
  • 🔍 Web Search — Admin-configurable web search integration
  • 🔔 Critical Error Tracking — Centralized error monitoring with email notifications
  • 📊 Configurable Embedding Dimensions — Support for Nova Multimodal Embeddings and flexible vector sizes

⚠️ Breaking Changes & Migration

AWS Parameter Store Migration

All shared environment variables have been migrated to AWS Parameter Store. Existing deployments must run the population script before deploying any services:

python3 scripts/populate_parameter_store.py --stage <stage> --dep-name <dep-name>

Services will fail to deploy without this step. See the Migration Guide for complete instructions.

Basic-Ops Service Elimination

The amplify-lambda-basic-ops service has been removed. Its /user-data functionality is now handled by amplify-lambda. If you have the amplify-lambda-basic-ops CloudFormation stack deployed, you must:

  1. Check if the stack exists: aws cloudformation describe-stacks --stack-name amplify-<dep-name>-basic-ops-<stage>
  2. If it exists, check for user storage data and handle the /user-data API Gateway endpoint conflict
  3. Follow Migration Guide Step 3a for detailed instructions

Coordinated Deployment Required

The backend and frontend must be deployed together — deploy the backend first, then rebuild and deploy the frontend container.

🆕 What's New

Model Updates

21 new models added, 1 updated, 3 removed (net +18):

Amazon Bedrock:

  • amazon.nova-2-multimodal-embeddings-v1:0 — Nova Multimodal Embeddings
  • amazon.nova-lite-v1:0, amazon.nova-micro-v1:0, amazon.nova-pro-v1:0 — Nova family

Google Gemini:

  • gemini-2.0-flash — Second-generation workhorse (1M context)
  • gemini-2.5-flash — Best price-performance with thinking features (1M context, 65K output)
  • gemini-2.5-pro — Promoted from preview to stable release

OpenAI GPT:

  • gpt-4.1-mini — Latest efficient model
  • gpt-5, gpt-5-mini, gpt-5.1, gpt-5.2 — Frontier reasoning and coding models

OpenAI Reasoning:

  • o3 — Advanced reasoning
  • o4-mini — Cost-effective compact reasoning

Anthropic Claude (Bedrock):

  • us.anthropic.claude-3-5-haiku-20241022-v1:0 — Claude 3.5 Haiku
  • us.anthropic.claude-opus-4-20250514-v1:0 — Claude 4 Opus
  • us.anthropic.claude-opus-4-5-20251101-v1:0 — Claude Opus 4.5
  • us.anthropic.claude-sonnet-4-20250514-v1:0 — Claude 4 Sonnet

Mistral:

  • us.mistral.pixtral-large-2502-v1:0 — Pixtral Large (124B multimodal)

Removed: amazon.titan-embed-text-v1, text-embedding-3-large, text-embedding-3-small (replaced by Nova embeddings)

Pull Request: #275

API Gateway Lambda Response Streaming

Migrated from Function URLs to API Gateway Lambda response streaming:

  • API Gateway streaming format with metadata JSON and 8-byte delimiter
  • Changed integration type from AWS to AWS_PROXY for streaming support
  • Updated Lambda invocation URI to use /2021-11-15/response-streaming-invocations endpoint

Pull Request: #275

Web Search Functionality

Admin-configurable web search integration:

  • Feature flag support for enabling/disabling web search
  • Migrated web search admin configuration to unified admin service
  • New webSearch.js module for search operations
  • Frontend preferences handling in assistants
  • Fixed web_search_preview tool triggering with images
  • Integration with OpenAI /v1/responses endpoint

Pull Request: #275

Critical Error Tracking & Monitoring

Centralized error monitoring infrastructure:

  • Critical error tracker service with database schema
  • Critical error processor for event handling
  • Email notification system (critical_error_notifier.py)
  • Integration across multiple Lambda services
  • New criticalLogger.js for JavaScript services
  • Schema validation and permissions for critical_errors endpoint

Pull Request: #275

Additional Charges & Billing

  • Support for additional charges tracking and billing
  • Custom cost attribution
  • Enhanced billing capabilities

Pull Request: #275

🔧 Embeddings & Vector Database

Nova Multimodal Embeddings Support

  • Amazon Bedrock Nova Multimodal Embeddings model support
  • Updated embedding code for Nova MME compatibility
  • Comprehensive documentation (EMBEDDING_DIMENSIONS.md, MODEL_CONFIGURATION.md)

Pull Requests: #264, #268

Configurable Embedding Dimensions

  • EMBEDDING_DIM environment variable (defaults to 1536)
  • Updated create_table.py for configurable vector dimensions
  • populate_parameter_store script updated with EMBEDDING_DIM format
  • Backward compatible: existing deployments unaffected (CREATE TABLE IF NOT EXISTS)
  • Cross-provider compatibility (Amazon Nova, Cohere Embed v3, OpenAI)

Pull Request: #264

Embedding Improvements

  • Enhanced embedding-dual-retrieval with better query handling
  • Dead Letter Queue (DLQ) processing for failed embedding chunks (embedding-dlq-handler.py)
  • Improved Excel file handling for corrupted files
  • Better error handling in visual transcription
  • Optimized database interactions
  • Enhanced logging for embedding diagnostics

Pull Requests: #262, #275

🔐 OAuth & Integrations

Microsoft Azure Admin Consent

  • New admin setting for Microsoft Azure integrations
  • Store consent settings in DynamoDB alongside existing integrations data
  • Check consent setting when user connects an integration

Pull Request: #266

OAuth Enhancements

  • Dynamic redirect URI support
  • Improved error handling and retry logic for consent errors
  • Better origin detection from event headers
  • Calendar time zone bug fixes

Pull Request: #275

💰 Billing & Cost Tracking

Usage Tracking

  • New centralized usageTracking.js module
  • Fixed null check for CachedCostPerThousandTokens in recordUsage
  • Support for OpenAI cached tokens in usage tracking
  • Handle response.completed usage format
  • New accounting.js for shared accounting logic

Pull Request: #265

Model Rate Table Updates

  • Added pricing for 21 new models
  • Updated model_rate_values.csv
  • Fixed Bedrock cost tracking bugs
  • Improved MTD cost calculations

Pull Requests: #265, #275

🐛 Bug Fixes

  • Bedrock validation errors: Convert tool roles to user, handle null toolUseId, disable reasoning with tools when incompatible (#265)
  • OpenAI usage format: Handle response.completed usage format and cached tokens in openaiUsageTransform (#265)
  • Visual transcription: Improved error handling and model retrieval logic (#262)
  • IAM policy deployment ordering: Fixed race condition with hardcoded ARN → !Ref for implicit CloudFormation dependencies
  • Lambda size limits: Added slimPatterns to strip 85MB Google API discovery cache from assistants-api layer

🤖 Assistants & Agent Framework

  • Refactored assistant API serverless configuration to use Python requirements layer
  • Enhanced user-defined assistants with better state management
  • Improved code interpreter integration
  • Added OpenAI provider support to agent framework
  • Enhanced agent prompt handling and core logic
  • Improved scheduled tasks processing
  • Removed legacy Docker build files and deployment scripts

Pull Request: #275

⚙️ Infrastructure & DevOps

Parameter Store Integration

  • All shared environment variables resolved from AWS SSM Parameter Store
  • populate_parameter_store.py script for initial population
  • Comprehensive migration guide (MIGRATION_README.md)

Var Template Updates

  • Updated <env>-var.yml.template with new embedding variables and comments
  • All dimensions remain default 1536 unless overridden

Pull Request: #268

Standalone Service Deployment

The following services are deployed independently (not included in serverless-compose.yml):

  • amplify-agent-loop-lambda/ — Agent Loop
  • amplify-lambda-assistants-api/ — Assistants API
  • amplify-lambda-assistants-api-office365/ — Office 365 Integration
  • amplify-lambda-assistants-api-google/ — Google Workspace Integration

📈 Stats

Metric Count
Commits 162
Files changed 358
Insertions +76,919
Deletions -49,986
Pull Requests 6
New AI Models 21

👥 Contributors

  • Karely Rodriguez (@karelyrodri) — Lead developer, core infrastructure, JIT provisioning, streaming
  • Allen Karns (@karnsab) — Release management, deployment fixes, parameter store migration
  • Jason Bradley (@jasonbrd) — Embedding dimensions, Nova MME, model updates, var template
  • Seviert (@seviert23) — Microsoft Azure admin consent integration

Deployment Instructions

  1. Run populate_parameter_store.py (required for all deployments)
  2. Handle basic-ops migration if applicable (see Migration Guide Step 3a)
  3. Deploy core services: serverless deploy --stage <env>
  4. Deploy standalone services as needed (see Step 5a in deployment docs)
  5. Rebuild and deploy the frontend container

See the full Migration Guide for detailed instructions.

Contributors

karelyrodri, jasonbrd, and 2 other contributors
Assets 2
Loading