|
1 | 1 | # Changelog
|
2 | 2 |
|
| 3 | +## 1.0.0 |
| 4 | + |
| 5 | +### ⚠ BREAKING CHANGES |
| 6 | + |
| 7 | +* Evaluation dataset schema is finalized with strong-type pydantic models. |
| 8 | + (previously saved eval file needs re-generation, for both adk eval cli and |
| 9 | + the eval tab in adk web UI). |
| 10 | +* `BuiltInCodeExecutor` (in code_executors package) replaces |
| 11 | + `BuiltInCodeExecutionTool` (previously in tools package). |
| 12 | +* All methods in services are now async, including session service, artifact |
| 13 | + service and memory service. |
| 14 | + * `list_events` and `close_session` methods are removed from session service. |
| 15 | +* agent.py file structure with MCP tools are now easier and simpler ([now](https://github.com/google/adk-python/blob/3b5232c14f48e1d5b170f3698d91639b079722c8/contributing/samples/mcp_stdio_server_agent/agent.py#L33) vs [before](https://github.com/google/adk-python/blob/a4adb739c0d86b9ae4587547d2653d568f6567f2/contributing/samples/mcp_agent/agent.py#L41)). |
| 16 | + Old format is not working anymore. |
| 17 | +* `Memory` schema and `MemoryService` is redesigned. |
| 18 | +* Mark various class attributes as private in the classes in the `tools` package. |
| 19 | +* Disabled session state injection if instruction provider is used. |
| 20 | + (so that you can have `{var_name}` in the instruction, which is required for code snippets) |
| 21 | +* Toolbox integration is revamped: tools/toolbox_tool.py → tools/toolbox_toolset.py. |
| 22 | +* Removes the experimental `remote_agent.py`. We'll redesign it and bring it back. |
| 23 | + |
| 24 | +### Features |
| 25 | + |
| 26 | +* Dev UI: |
| 27 | + * A brand new trace view for overall agent invocation. |
| 28 | + * A revamped evaluation tab and comparison view for checking eval results. |
| 29 | +* Introduced `BaseToolset` to allow dynamically add/remove tools for agents. |
| 30 | + * Revamped MCPToolset with the new BaseToolset interface. |
| 31 | + * Revamped GoogleApiTool, GoogleApiToolset and ApplicationIntegrationToolset with the new BaseToolset interface. |
| 32 | + * Resigned agent.py file structure when needing MCPToolset. |
| 33 | + * Added ToolboxToolset. |
| 34 | +* Redesigned strong-typed agent evaluation schema. |
| 35 | + * Allows users to create more cohesive eval sets. |
| 36 | + * Allows evals to be extended for non-text modality. |
| 37 | + * Allows for a structured interaction with the uber eval system. |
| 38 | +* Redesigned Memory schema and MemoryService interfaces. |
| 39 | +* Added token usage to LlmResponse. |
| 40 | +* Allowed specifying `--adk_version` in `adk deploy cloud_run` cli. Default is the current version. |
| 41 | + |
| 42 | +### Bug Fixes |
| 43 | + |
| 44 | +* Fixed `adk deploy cloud_run` failing bug. |
| 45 | +* Fixed logs not being printed due to `google-auth` library. |
| 46 | + |
| 47 | +### Miscellaneous Chores |
| 48 | + |
| 49 | +* Display full help text when adk cli receives invalid arguments. |
| 50 | +* `adk web` now binds `127.0.0.1` by default, instead of 0.0.0.0. |
| 51 | +* `InMemoryRunner` now takes `BaseAgent` in constructor. |
| 52 | +* Various docstring improvements. |
| 53 | +* Various UI tweaks. |
| 54 | +* Various bug fixes. |
| 55 | +* Update various contributing/samples for contributors to validate the implementation. |
| 56 | + |
| 57 | + |
3 | 58 | ## 0.5.0
|
4 | 59 |
|
5 | 60 | ### ⚠ BREAKING CHANGES
|
|
0 commit comments