* fix: update state key references
* update docs to include info on state injection in instructions
* update java snippet
* add docs on InstructionProvider
* add tooltip on direct state modification
---------
Co-authored-by: Lavi Nigam <98014943+lavinigam-gcp@users.noreply.github.com>
Updated the Markdown link to point directly to the relevant section (output_key) within the LLM Agent documentation, instead of linking to the top of the page.
Co-authored-by: Lavi Nigam <98014943+lavinigam-gcp@users.noreply.github.com>
- #254 docs/sessions/session.md -> …part of session object's identification.
In the session object identification description, it was unclear (for first time readers), what exactly is id, and app_name and how they are different. Added some examples according to my own understanding. Please change accordingly if you feel so. Thanks.
- #252 Multiple files: quickstart-streaming.md, quickstart.md, mcp-tools.md
add an option in the CLI to enable or disable the reload feature. So users(esp. windows) can disable this if they come across the '_make_subprocess_transport NotImplementedError' bug on windows.
- #179 docs/agents/models.md -> Add Windows Encoding Note for LiteLLM
When using ADK agents with LiteLlm on Windows, users might encounter a UnicodeDecodeError: 'charmap' codec can't decode byte... error. This often happens when litellm (used by LiteLlm) tries to read cached files (like model pricing information) using the default Windows encoding (cp1252) instead of UTF-8.
- #170 docs/tools/mcp-tools.md fix: update ADK setup link and add .env file location note in MCPTools.
Also resolved the issue: #169
* feat: Improve workflow agent examples (Loop, Parallel, Sequence)
This commit enhances the examples for LoopAgent, ParallelAgent, and SequenceAgent to provide clearer usage patterns, demonstrate key features, and improve developer experience.
Agent-Specific Fixes & Enhancements:
LoopAgent:
Adds an example demonstrating CheckCondition usage, showing both max_iterations and custom condition-based termination.
Fixesgoogle/adk-docs#67, google/adk-docs#148, google/adk-python#315
ParallelAgent:
Adds an example showing how to merge outputs from parallel branches using a subsequent SequentialAgent and agent states.
Demonstrates state/output control after parallel execution.
Fixesgoogle/adk-docs#142
SequenceAgent:
Clarifies the example to show explicit use of agent states within instructions for passing outputs between sequential steps.
Fixesgoogle/adk-docs#150
General Example Improvements (Applied across all three):
Updated examples to use InMemoryRunner.
Adopted new, clearer instruction prompt templates.
Included runner.run_async calls in standalone .py files to demonstrate programmatic execution outside adk web.
Enhanced InMemoryRunner print output for better visibility of individual agent outputs during execution.
Added commented-out asyncio.run() calls in .py files for environments requiring explicit async loop management.
Included multiple example queries in .py files for testing.
Added "Part of agent.py -->" comment for easy copy-pasting into agent.py for use with adk web.
Added copyright and license headers for each file.
* fix: instructions for LoopAgent is cleaned up and removed backticks. Changed the init to include initial state in the ---8<
* feat(LoopAgent): add exit_loop tool for controlled exit
Introduces the `exit_loop` tool, enabling the agent to programmatically trigger an exit from its processing loop based on internal logic or tool usage.
fix(LoopAgent): correct state f-string formatting
Resolves runtime errors caused by incorrect f-string formatting when state variables were used within the agent's execution flow.
chore: Temporarily comment out session/runner setup
Comments out initialization code related to session, runner, and agent interaction for debugging/refactoring purposes. This section should be revisited.
Update the variable name in `docs/agents/custom-agents.md`.
* Change the variable name to `GEMINI_2_FLASH` on line 114 to maintain consistency with example's code
Co-authored-by: Kristopher Overholt <koverholt@gmail.com>
* fix: update gemini 2.0 flash to latest model id for Live API
* fix: use -flash for most pages instead of the live api model
* fix: wording andformatting
* fix: resolve merge conflict
Sita Lakshmi Sangameswaran