Skip to content

Releases: aixplain/aiXplain

0.2.43

11 Mar 20:12

Choose a tag to compare

Release Notes

0.2.43 (2026-03-11)

Features

  • v1-restructure: Reorganize legacy modules into aixplain/v1/ namespace (#859)

    • All v1 modules (enums, factories, modules, decorators, base, processes) moved from aixplain/ to aixplain/v1/
    • Backward-compatible import redirector (aixplain/_compat.py) ensures all existing import paths (from aixplain.modules import …, from aixplain.factories import …, etc.) continue to work transparently
    • Clean separation between v1 and v2 SDK surfaces
  • agent: Add run_response_generation parameter support for v2 agents

    • Snake-case run_response_generation kwarg auto-mapped to runResponseGeneration
    • Available on both run() and run_async() methods
  • api-key: Improved v2 API key management (#851, #853)

    • model_id renamed to model across APIKeyLimits and APIKeyUsageLimit for consistency
    • access_key renamed to key on APIKey for a cleaner interface
    • Accept model objects (not just IDs) when setting rate limits — auto-resolved via _resolve_model()
    • before_save() deduplicates by name, falling back to update if a key with the same name exists
    • _create() gracefully handles HTTP 422 name conflicts
    • User-facing to_dict() now returns snake_case keys

Improvements

  • agent: Updated default LLM model ID for v2 agents
  • agent: Fixed duplicate docstring in Agent.run() method
  • agent: Task dependencies field no longer excluded when empty
  • v2-model: Refined model and resource modules with improved error handling
  • v2-integration: Updated integration module improvements
  • readme: Rewrote README with v2-first examples, AgenticOS architecture diagram, and streamlined quick-start guide
  • docs: Regenerated API documentation for new v1/v2 structure (#845)
  • dependencies: Updated requirements and lockfile

Bug Fixes

  • tests: Fixed test repetition and revised merge conflicts (#859, #860)
  • pipeline: Removed duplicate code in pipeline module
  • v2: Various cleanup across v2 exceptions, client, and utility modules

Test Changes

  • Added v2 agent functional tests
  • Expanded v2 model functional tests with score threshold tests
  • Refactored agent, team agent, and inspector functional tests
  • Updated unit tests for v2 API key and model changes

0.2.42

03 Mar 21:09

Choose a tag to compare

Release Notes

0.2.42 (2026-03-04)

Features

  • v2-tool: Add as_tool() method to V2 Tool class (#834)

    • Extends Model.as_tool() to include allowed_actions in serialization
    • Ensures agents only receive the permitted set of actions from a tool
  • index-model: Support URI-based text records for index upsert (#834)

    • Text records now accept either a value or a uri (previously only value)
    • Non-image file URIs are no longer rejected, enabling document indexing

Improvements

  • model-cache: Graceful fallback on cache failures (#834)

    • Cache lookup errors now log a warning and fall back to direct API fetch
    • Model retrieval no longer breaks due to transient caching issues
  • team-agent: Parallel sub-agent fetching with model cache (#834)

    • Sub-agents fetched concurrently via ThreadPoolExecutor
    • Model cache avoids duplicate lookups of the same model ID
  • team-agent: Updated default serper search model ID in YAML builder

Bug Fixes

  • model-cache: Fix error propagation when cache and list fetch both fail (#834)
    • Previously raised raw exception; now falls back to direct _fetch_model_by_id

Test Changes

  • Removed deprecated finetune functional tests and CI workflows (finetune_v1, finetune_v1_model, finetune_v2)
  • Added V2 tool functional tests
  • Added model run functional tests
  • Refactored agent, team agent, evolver, and inspector functional tests
  • Simplified test deletion utilities
  • Updated index model unit tests for new value/uri validation

0.2.41

17 Feb 21:10

Choose a tag to compare

Release Notes

0.2.41 (2026-02-18)

Features

  • api-key: Add APIKey resource to SDK v2 (#810)

    • Full API key management support in v2 module
    • Create, retrieve, update, and delete API keys programmatically
  • model: Add streaming support for v2 models (#816)

    • Real-time streaming responses for model inference
    • Improved latency for long-running model calls
  • meta-agents: Add Debugger meta-agent for agent response analysis (#814)

    • Standalone usage: aix.Debugger().run(...)
    • Chained from agent response: response.debug()
  • model: Add get_models_by_name functionality (#674)

    • Search and retrieve models by their display name
    • Simplified model discovery workflow
  • agent: Adjust run methods so that run() uses the v2 endpoint (#817)

    • run() now defaults to v2 API
    • run_async() available for asynchronous execution

Improvements

  • inspector: Updated inspector payload structure for better compatibility (#815)
  • model: Make model tool function, supplier, and version fields optional (#813)
  • agent: Removed limitations other than abort-critical for more flexible execution (#811)
  • inspector: Remove INFO severity logging from inspector for cleaner output
  • docs: Regenerate API documentation (#820)

Bug Fixes

  • model: Fix models API response handling (#822)
  • v2: Fix code/value attribute handling for actions (#808)
  • v2: Fix v2 imports from v1 modules that trigger env var validation (#819)
  • v2: Remove duplicate code in v2 modules (HOTFIX)

0.2.40

30 Jan 18:20
d4d17dc

Choose a tag to compare

0.2.40 (2026-01-30)

Features

  • agent: Add real-time progress tracking and display for agent execution (#818)
    • New agent_progress module with terminal and Jupyter/Colab notebook support
    • Configurable display formats: "status" (single line) and "logs" (timeline)
    • Verbosity levels 1 (minimal), 2 (thoughts), 3 (full I/O)
    • Replaces old show_progress boolean with progress_format, progress_verbosity, and progress_truncate parameters
  • agent: Add variables parameter to agent run for {{variable}} placeholder substitution in instructions and description (#818)
  • agent: Add runResponseGeneration parameter to agent run (#818)
  • meta-agents: New Debugger meta-agent for analyzing agent responses (#818)
    • Standalone usage: aix.Debugger().run(...)
    • Chained from agent response: response.debug()
  • api-key: Add TokenType enum (INPUT, OUTPUT, TOTAL) for granular token-based rate limiting (#818)
    • New token_type parameter on APIKeyLimits
  • params: Extend Parameter with is_fixed, values, default_values, available_options, data_type, data_sub_type, and multiple_values fields (#818)
  • inspector: Add EDIT action type to inspector actions (#818)

Improvements

  • inspector: Refactor and simplify inspector implementation in both v1 and v2 (#818)
    • Migrated v1 inspector to Pydantic models
    • Streamlined v2 inspector API
  • resource: Consolidate asset_path and instance_id into a single path field (priority: instanceId > assetPath) (#818)
  • agent: Rename intermediate_steps to steps in agent response data (#818)
  • agent: Add session_id to AgentResponseData (#818)
  • resource: Fix before_run / after_run hook invocation to avoid double calls when run() delegates to run_async() (#818)
  • tool: Add subscriptions field to Tool (#818)
  • tool: Remove synthetic "run" action fallback for utility models (#818)
  • integration: Code formatting and simplification (#818)

Bug Fixes

  • resource: Fix execution_stats field mapping to use executionStats (#818)

Removals

  • Removed inspector_factory.py from team agent factory (#818)
  • Removed auth_scheme field from Tool (#818)
  • Removed deprecated agent factory utils unit tests (#818)
  • Removed old inspector functional tests (replaced with new v2 tests) (#818)

Tests

  • Added test_client.py - comprehensive v2 client unit tests (#818)
  • Added test_exceptions.py - v2 exception handling unit tests (#818)
  • Added test_inspector.py - v2 inspector unit tests (#818)
  • Added test_meta_agents.py - meta-agents unit tests (#818)
  • Expanded test_resource.py and test_core.py significantly (#818)
  • Added tests/functional/apikey/test_api.py for API key functional tests (#818)
  • Added tests/functional/v2/inspector_functional_test.py (#818)
  • Updated agent, model, tool, and integration functional tests (#818)

0.2.39

18 Dec 20:27
6e076b1

Choose a tag to compare

0.2.39 (2025-12-18)

Features

  • script-tools: Migrate to new script connection tool creation approach (#741)
    • New ModelFactory.create_script_connection_tool() method
    • Updated ToolFactory.create() with new approach
    • Deprecated ModelFactory.create_utility_model() and removed AgentFactory.create_custom_python_code_tool()
  • agent: Add inspector functionality to build agent (#704)
  • agent: Add steps to agent response data (#732)
  • model: Add response format support to model run (#692)
  • model: Add trace_request parameter for debugging (#686)
  • model: Streaming progress improvements (#701)
  • integration: Add description field to integration create (#739)
  • docs: Automate documentation generation (#710)

Bug Fixes

  • model: Fix ResponseStatus string-to-enum conversion (#761)
  • model: Fix cache storage consistency in AssetCache.add_list() (#761)
  • mcp: Make authentication_schema optional for MCP connections (#761)
  • script-tools: Fix nested function support in Python sandbox (#758)
  • script-tools: Fix auto-detection of functions from provided code (#758)
  • validation: Centralize API key validation and fix misleading error messages (#684)
  • enum: Fix supplier enum ID handling (#757)
  • params: Fix action params reading (#753)
  • params: Make temperature/top_p optional (#705)
  • params: Fix time default values (#708)
  • utils: Fix data handling for empty objects (#751)
  • error: Improve error propagation (#712)
  • model: Normalize expected output from BaseModel (#685)
  • tests: Fix v2 tests by removing v2 references (#736)
  • tests: Fix agent functional tests (#759)
  • tests: Mark flaky test_specific_model_parameters_e2e (#720)

Code Quality

  • Move deprecated params to kwargs (#699)
  • Add module docstrings to fix linting errors (#761)
  • Fix issubclass() TypeError (#705)
  • Enhanced error handling and retry logic in model utils (#761)

Deprecated

  • ModelFactory.create_utility_model() - Use ModelFactory.create_script_connection_tool() instead
  • AgentFactory.create_custom_python_code_tool() - Removed, use ToolFactory.create() instead

Dependencies

  • pydantic: >=2.10.6
  • Jinja2: 3.1.6
  • Python: >=3.9, <4

0.2.38

10 Nov 18:02
7c0e80a

Choose a tag to compare

0.2.38 Pre-release
Pre-release

What's Changed

Full Changelog: 0.2.36...0.2.38

0.2.37

05 Nov 19:11
f7ce3bb

Choose a tag to compare

Release Notes: aiXplain SDK v0.2.37

Release Date: November 5, 2025

🎉 What's New

🚀 Real-Time Progress Streaming

  • Agent Progress Display: Added inline progress display for agent and team agent execution (#701)
    • New show_progress parameter for Agent.run() and TeamAgent.run() methods
    • Real-time progress updates showing stage, tool, runtime, success status, and reason
    • Progress bar for team agents displaying current_step/total_steps
    • Detailed progress format showing tool input/output and full execution details

🔍 Enhanced Inspector System

  • Inspector Improvements: Major updates to the Inspector system for team agents (#716, #704)
    • Inspector changed to default None and renamed for better clarity
    • Added inspector integration in agent building process
    • Enhanced inspector functionality with better error propagation (#712)
    • Comprehensive inspector tests and validation (#706)

📦 Parameter Management & Deprecation Handling

  • Deprecated Parameters Migration: Moved deprecated parameters to kwargs for cleaner API (#699)
    • Better handling of legacy parameters in Agent and TeamAgent creation
    • Improved parameter validation and deprecation warnings
    • Enhanced backward compatibility while encouraging modern usage

🔧 Core Improvements

  • Error Propagation: Improved error handling and propagation throughout the system (#712)
  • Time Handling: Fixed time default values for better consistency (#708)
  • Agent Warning Messages: Modified agent warning messages for better user experience (#707)
  • Temperature/Top_p Parameters: Made temperature and top_p parameters optional to prevent TypeError issues (#705)

🐛 Bug Fixes

  • isSubclass TypeError: Fixed TypeError in issubclass() by adding proper class checks (#705)
  • Merge Conflicts: Resolved various merge conflicts and test issues (#714)
  • Test Stability: Fixed non-inspector related tests and improved test reliability (#709)
  • Literal Instructions: Fixed literal instructions issue for better text handling

📚 Documentation & Developer Experience

  • Automated Documentation Generation: Added automated docs generation workflow (#710)
    • Streamlined documentation build process
    • Updated docs.yaml configuration for better automation
  • Enhanced Testing: Comprehensive inspector tests and improved test coverage (#706)
  • Code Quality: Continued improvements to code quality and test reliability

📦 Installation & Upgrade

Install or upgrade to the latest version:

pip install --upgrade aiXplain

Or install a specific version:

pip install aiXplain==0.2.37

🔧 Breaking Changes

None - This release maintains backward compatibility with existing code.

🎯 Key Features Spotlight

Real-Time Progress Streaming

from aixplain.factories import AgentFactory, TeamAgentFactory

# Agent with progress display
agent = AgentFactory.create(name="MyAgent", ...)
response = agent.run(
    query="Analyze this data",
    progress_verbosity='compact'  # 'compact', 'full' or None, default to 'compact'
)

# Team agent with progress bar
team_agent = TeamAgentFactory.create(name="MyTeam", agents=[...])
response = team_agent.run(
    query="Complex multi-step task",
    progress_verbosity='compact'  # 'compact', 'full' or None, default to 'compact'
)

Enhanced Inspector System

from aixplain.modules.team_agent import InspectorTarget
from aixplain.modules.team_agent.inspector import Inspector, InspectorOutput, InspectorAction, InspectorAuto, VerificationInspector
from aixplain.modules.model.response import ModelResponse


team = TeamAgentFactory.create(
    name="Audio Sentiment Team 3",
    description="Converts speech to text and analyzes its sentiment.",
    agents=[text_analysis_agent, multimedia_agent],
    llm_id="677c16166eb563bb611623c1",  # Llama 3.3
    inspectors=[VerificationInspector()]
)

👥 Contributors

Special thanks to all contributors who made this release possible:

  • @ahmetgunduz - Progress streaming, inspector improvements, and core fixes
  • @elsheikhams99 - Deprecated parameters migration and API improvements
  • @ikxplain - Testing and integration work
  • And all other contributors who helped with testing, bug reports, and feedback

📖 Full Changelog

Full Changelog: 0.2.36...0.2.37


🆘 Need Help?

0.2.36

10 Oct 14:38
3494aa9

Choose a tag to compare

Changelog

All notable changes to the aiXplain SDK will be documented in this file.

The format is based on Keep a Changelog,
and this project adheres to Semantic Versioning.

[0.2.36] - 2025-10-10

🚀 Features

  • Agent & Tool Deployment: Automatic deployment of tools and agents with MCP integration (#558, #680)
    • Added MCP deploy method for seamless agent deployment
    • Enhanced tool and agent automation capabilities

🐛 Bug Fixes

  • Model Pagination: Fixed pagination issue when fetching models by IDs (#643)
  • Utility Function: Resolved input parsing bug in utility functions (#648)
  • AgentFactory: Clear saved information properly in AgentFactory to prevent state leakage (#668)
  • MCP Deployment: Fixed missing MCP deploy method (#680)

📚 Documentation

  • API Reference: Updated and improved API reference documentation (#672)
  • Docstrings: Added missing docstrings for evolver module (#671)
  • Documentation Site: Added comprehensive docs structure (#641)

🔧 Development & Testing

  • Pre-commit Configuration: Added Ruff linter pre-commit hooks for better code quality (#670, #671)
  • Test Improvements:
    • Removed model deletion from tests for better reliability (#644)
    • Added time delay before searching indexes to improve test stability (#646)
  • Logging: Updated lockfile logs for better debugging (#666)
  • CI/CD: Updated GitHub Actions workflow configuration (#640)

👥 Contributors

Special thanks to all contributors who made this release possible:

Full Changelog: 0.2.35...0.2.36


How to Upgrade

To upgrade to the latest version:

pip install --upgrade aiXplain

For specific version:

pip install aiXplain==0.2.36

Breaking Changes

No breaking changes in this release.

Deprecation Notices

No deprecations in this release.

0.2.35

09 Sep 18:43
6b86d30

Choose a tag to compare

What's Changed

Full Changelog: 0.2.34...0.2.35

0.2.34

28 Aug 19:25
632b4b3

Choose a tag to compare

What's Changed

New Contributors

Full Changelog: 0.2.33...0.2.34