
* update readme * update * update * update * update * update * update * add sample project * revert notebook change back * update * update interactive version * add nuget package * refactor Message * update example * add azure nightly build pipeline * Set up CI with Azure Pipelines [skip ci] * Update nightly-build.yml for Azure Pipelines * add dotnet interactive package * add dotnet interactive package * update pipeline * add nuget feed back * remove dotnet-tool feed * remove dotnet-tool feed comment * update pipeline * update build name * Update nightly-build.yml * Delete .github/workflows/dotnet-ci.yml * update * add working_dir to use step * add initateChat api * update oai package * Update dotnet-build.yml * Update dotnet-run-openai-test-and-notebooks.yml * update build workflow * update build workflow * update nuget feed * update nuget feed * update aoai and sk version * Update InteractiveService.cs * add support for GPT 4V * add DalleAndGPT4V example * update example * add user proxy agent * add readme * bump version * update example * add dotnet interactive hook * update * udpate tests * add website * update index.md * add docs * update doc * move sk dependency out of core package * udpate doc * Update Use-function-call.md * add type safe function call document * update doc * update doc * add dock * Update Use-function-call.md * add GenerateReplyOptions * remove IChatLLM * update version * update doc * update website * add sample * fix link * add middleware agent * clean up doc * bump version * update doc * update * add Other Language * remove warnings * add sign.props * add sign step * fix pipelien * auth * real sign * disable PR trigger * update * disable PR trigger * use microbuild machine * update build pipeline to add publish to internal feed * add internal feed * fix build pipeline * add dotnet prefix * update ci * add build number * update run number * update source * update token * update * remove adding source * add publish to github package * try again * try again * ask for write pacakge * disable package when branch is not main * update * implement streaming agent * add test for streaming function call * update * fix #1588 * enable PR check for dotnet branch * add website readme * only publish to dotnet feed when pushing to dotnet branch * remove openai-test-and-notebooks workflow * update readme * update readme * update workflow * update getting-start * upgrade test and sample proejct to use .net 8 * fix global.json format && make loadFromConfig API internal only before implementing * update * add support for LM studio * add doc * Update README.md * add push and workflow_dispatch trigger * disable PR for main * add dotnet env * Update Installation.md * add nuget * refer to newtonsoft 13 * update branch to dotnet in docfx * Update Installation.md * pull out HumanInputMiddleware and FunctionCallMiddleware * fix tests * add link to sample folder * refactor message * refactor over IMessage * add more tests * add more test * fix build error * rename header * add semantic kernel project * update sk example * update dotnet version * add LMStudio function call example * rename LLaMAFunctin * remove dotnet run openai test and notebook workflow * add FunctionContract and test * update doc * add documents * add workflow * update * update sample * fix warning in test * reult length can be less then maximumOutputToKeep (#1804) * merge with main * add option to retrieve inner agent and middlewares from MiddlewareAgent * update doc * adjust namespace * update readme * fix test * use IMessage * more updates * update * fix test * add comments * use FunctionContract to replace FunctionDefinition * move AutoGen contrac to AutoGen.Core * update installation * refactor streamingAgent by adding StreamingMessage type * update sample * update samples * update * update * add test * fix test * bump version * add openaichat test * update * Update Example03_Agent_FunctionCall.cs * [.Net] improve docs (#1862) * add doc * add doc * add doc * add doc * add doc * add doc * update * fix test error * fix some error * fix test * fix test * add more tests * edits --------- Co-authored-by: ekzhu <ekzhu@users.noreply.github.com> * [.Net] Add fill form example (#1911) * add form filler example * update * fix ci error * [.Net] Add using AutoGen.Core in source generator (#1983) * fix using namespace bug in source generator * remove using in sourcegenerator test * disable PR test * Add .idea to .gitignore (#1988) * [.Net] publish to nuget.org feed (#1987) * publish to nuget * update ci * update dotnet-release * update release pipeline * add source * remove empty symbol package * update pipeline * remove tag * update installation guide * [.Net] Rename some classes && APIs based on doc review (#1980) * rename sequential group chat to round robin group chat * rename to sendInstruction * rename workflow to graph * rename some api * bump version * move Graph to GroupChat folder * rename fill application example * [.Net] Improve package description (#2161) * add discord link and update package description * Update getting-start.md * [.Net] Fix document comment from the most recent AutoGen.Net engineer sync (#2231) * update * rename RegisterPrintMessageHook to RegisterPrintMessage * update website * update update.md * fix link error * [.Net] Enable JsonMode and deterministic output in AutoGen.OpenAI OpenAIChatAgent (#2347) * update openai version && add sample for json output * add example in web * update update.md * update image url * [.Net] Add AutoGen.Mistral package (#2330) * add mstral client * enable streaming support * add mistralClientAgent * add test for function call * add extension * add support for toolcall and toolcall result message * add support for aggregate message * implement streaming function call * track (#2471) * [.Net] add mistral example (#2482) * update existing examples to use messageCOnnector * add overview * add function call document * add example 14 * add mistral token count usage example * update version * Update dotnet-release.yml (#2488) * update * revert gitattributes --------- Co-authored-by: mhensen <mh@webvize.nl> Co-authored-by: ekzhu <ekzhu@users.noreply.github.com> Co-authored-by: Krzysztof Kasprowicz <60486987+Krzysztof318@users.noreply.github.com>
2.7 KiB
Overview of function call
In some LLM models, you can provide a list of function definitions to the model. The function definition is usually essentially an OpenAPI schema object which describes the function, its parameters and return value. And these function definitions tells the model what "functions" are available to be used to resolve the user's request. This feature greatly extend the capability of LLM models by enabling them to "execute" arbitrary function as long as it can be described as a function definition.
Below is an example of a function definition for getting weather report for a city:
Note
To use function call, the underlying LLM model must support function call as well for the best experience. The model used in the example below is
gpt-3.5-turbo-0613
.
{
"name": "GetWeather",
"description": "Get the weather report for a city",
"parameters": {
"city": {
"type": "string",
"description": "The city name"
},
"required": ["city"]
},
}
When the model receives a message, it will intelligently decide whether to use function call or not based on the message received. If the model decides to use function call, it will generate a function call which can be used to invoke the actual function. The function call is a json object which contains the function name and its arguments.
Below is an example of a function call object for getting weather report for Seattle:
{
"name": "GetWeather",
"arguments": {
"city": "Seattle"
}
}
And when the function call is return to the caller, it can be used to invoke the actual function to get the weather report for Seattle.
Create type-safe function contract and function call wrapper use AutoGen.SourceGenerator
AutoGen provides a source generator to easness the trouble of manually craft function contract and function call wrapper from a function. To use this feature, simply add the AutoGen.SourceGenerator
package to your project and decorate your function with Function
attribute.
For more information, please check out Create type-safe function.
Use function call in an agent
AutoGen provides first-class support for function call in its agent story. Usually there are three ways to enable a function call in an agent.
- Pass function definitions when creating an agent. This only works if the agent supports pass function call from its constructor.
- Passing function definitions in @AutoGen.Core.GenerateReplyOptions when invoking an agent
- Register an agent with @AutoGen.Core.FunctionCallMiddleware to process and invoke function calls.
For more information, please check out Use function call in an agent.