agentevals 0.0.3 → 0.0.5

package/README.md

@@ -9,7 +9,7 @@ It is intended to provide a good conceptual starting point for your agent's eval
 
 If you are looking for more general evaluation tools, please check out the companion package [`openevals`](https://github.com/langchain-ai/openevals).
 
-## Quickstart
+# Quickstart
 
 To get started, install `agentevals`:
 
@@ -72,17 +72,20 @@ console.log(evalResult);
 }
 ```
 
-You can see that despite the small difference in the final response and tool calls, the evaluator still returns a score of `true` since the overall trajectory is the same between the output and reference!
+You can see that the evaluator returns a score of `true` since the overall trajectory is a reasonable path for the agent to take to answer the user's question.
 
-## Table of Contents
+For more details on this evaluator, including how to customize it, see the section on [trajectory LLM-as-judge](#trajectory-llm-as-judge).
+
+# Table of Contents
 
 - [Installation](#installation)
 - [Evaluators](#evaluators)
-  - [Agent Trajectory](#agent-trajectory)
+  - [Agent Trajectory Match](#agent-trajectory-match)
     - [Strict match](#strict-match)
     - [Unordered match](#unordered-match)
     - [Subset/superset match](#subset-and-superset-match)
-    - [Trajectory LLM-as-judge](#trajectory-llm-as-judge)
+    - [Tool args match modes](#tool-args-match-modes)
+  - [Trajectory LLM-as-judge](#trajectory-llm-as-judge)
   - [Graph Trajectory](#graph-trajectory)
     - [Graph trajectory LLM-as-judge](#graph-trajectory-llm-as-judge)
     - [Graph trajectory strict match](#graph-trajectory-strict-match)
@@ -91,7 +94,7 @@ You can see that despite the small difference in the final response and tool cal
   - [Pytest or Vitest/Jest](#pytest-or-vitestjest)
   - [Evaluate](#evaluate)
 
-## Installation
+# Installation
 
 You can install `agentevals` like this:
 
@@ -106,44 +109,70 @@ npm install openai
 ```
 
 It is also helpful to be familiar with some [evaluation concepts](https://docs.smith.langchain.com/evaluation/concepts) and
-LangSmith's Vitest/Jest integration for running evals, which is documented [here](https://docs.smith.langchain.com/evaluation/how_to_guides/pytest).
+LangSmith's pytest integration for running evals, which is documented [here](https://docs.smith.langchain.com/evaluation/how_to_guides/pytest).
 
-## Evaluators
+# Evaluators
 
-### Agent trajectory
+## Agent trajectory match
 
-Agent trajectory evaluators are used to judge the trajectory of an agent's execution either against an expected trajectory or using an LLM.
+Agent trajectory match evaluators are used to judge the trajectory of an agent's execution either against an expected trajectory or using an LLM.
 These evaluators expect you to format your agent's trajectory as a list of OpenAI format dicts or as a list of LangChain `BaseMessage` classes, and handle message formatting
 under the hood.
 
-#### Strict match
+AgentEvals offers the `create_trajectory_match_evaluator`/`createTrajectoryMatchEvaluator` and `create_async_trajectory_match_evaluator` methods for this task. You can customize their behavior in a few ways:
+
+- Setting `trajectory_match_mode`/`trajectoryMatchMode` to [`strict`](#strict-match), [`unordered`](#unordered-match), [`subset`](#subset-and-superset-match), or [`superset`](#subset-and-superset-match) to provide the general strategy the evaluator will use to compare trajectories
+- Setting [`tool_args_match_mode`](#tool-args-match-modes) and/or [`tool_args_match_overrides`](#tool-args-match-modes) to customize how the evaluator considers equality between tool calls in the actual trajectory vs. the reference. By default, only tool calls with the same arguments to the same tool are considered equal.
+
+### Strict match
 
-The `trajectory_strict_match` evaluator, compares two trajectories and ensures that they contain the same messages
-in the same order with the same tool calls. It allows for differences in message content and tool call arguments,
-but requires that the selected tools at each step are the same.
+The `"strict"` `trajectory_match_mode` compares two trajectories and ensures that they contain the same messages
+in the same order with the same tool calls. Note that it does allow for differences in message content:
 
 ```ts
-import { trajectoryStrictMatch } from "agentevals";
+import { createTrajectoryMatchEvaluator } from "agentevals";
 
 const outputs = [
-    { role: "user", content: "What is the weather in SF?" },
-    {
-      role: "assistant",
-      tool_calls: [{
-        function: { name: "get_weather", arguments: JSON.stringify({ city: "SF" }) }
-      }]
-    },
-    { role: "tool", content: "It's 80 degrees and sunny in SF." },
-    { role: "assistant", content: "The weather in SF is 80 degrees and sunny." },
+  { role: "user", content: "What is the weather in SF?" },
+  {
+    role: "assistant",
+    content: "",
+    tool_calls: [{
+      function: {
+        name: "get_weather",
+        arguments: JSON.stringify({ city: "San Francisco" })
+      },
+    }, {
+      function: {
+        name: "accuweather_forecast",
+        arguments: JSON.stringify({"city": "San Francisco"}),
+      },
+    }]
+  },
+  { role: "tool", content: "It's 80 degrees and sunny in SF." },
+  { role: "assistant", content: "The weather in SF is 80 degrees and sunny." },
 ];
 
 const referenceOutputs = [
-    { role: "user", content: "What is the weather in San Francisco?" },
-    { role: "assistant", tool_calls: [{ function: { name: "get_weather", arguments: JSON.stringify({ city: "San Francisco" }) } }] },
-    { role: "tool", content: "It's 80 degrees and sunny in San Francisco." },
+  { role: "user", content: "What is the weather in San Francisco?" },
+  {
+    role: "assistant",
+    content: "",
+    tool_calls: [{
+      function: {
+        name: "get_weather",
+        arguments: JSON.stringify({ city: "San Francisco" })
+      }
+    }]
+  },
+  { role: "tool", content: "It's 80 degrees and sunny in San Francisco." },
 ];
 
-const result = await trajectoryStrictMatch({
+const evaluator = createTrajectoryMatchEvaluator({
+  trajectoryMatchMode: "strict",
+})
+
+const result = await evaluator({
   outputs,
   referenceOutputs,
 });
@@ -153,22 +182,27 @@ console.log(result);
 
 ```
 {
-    'key': 'trajectory_accuracy',
-    'score': true,
+    'key': 'trajectory_strict_match',
+    'score': false,
 }
 ```
 
-#### Unordered match
+`"strict"` is useful is if you want to ensure that tools are always called in the same order for a given query (e.g. a company policy lookup tool before a tool that requests vacation time for an employee).
+
+**Note:** If you would like to configure the way this evaluator checks for tool call equality, see [this section](#tool-args-match-modes).
 
-The `trajectory_unordered_match` evaluator, compares two trajectories and ensures that they contain the same number of tool calls in any order. This is useful if you want to allow flexibility in how an agent obtains the proper information, but still do care that all information was retrieved.
+### Unordered match
+
+The `"unordered"` `trajectory_match_mode` compares two trajectories and ensures that they contain the same tool calls in any order. This is useful if you want to allow flexibility in how an agent obtains the proper information, but still do care that all information was retrieved.
 
 ```ts
-import { trajectoryUnorderedMatch } from "agentevals";
+import { createTrajectoryMatchEvaluator } from "agentevals";
 
 const outputs = [
   { role: "user", content: "What is the weather in SF and is there anything fun happening?" },
   {
     role: "assistant",
+    content: "",
     tool_calls: [{
       function: {
         name: "get_weather",
@@ -179,6 +213,7 @@ const outputs = [
   { role: "tool", content: "It's 80 degrees and sunny in SF." },
   {
     role: "assistant",
+    content: "",
     tool_calls: [{
       function: {
         name: "get_fun_activities",
@@ -194,6 +229,7 @@ const referenceOutputs = [
   { role: "user", content: "What is the weather in SF and is there anything fun happening?" },
   {
     role: "assistant",
+    content: "",
     tool_calls: [
       {
         function: {
@@ -214,7 +250,11 @@ const referenceOutputs = [
   { role: "assistant", content: "In SF, it's 80˚ and sunny, but there is nothing fun happening." },
 ];
 
-const result = await trajectoryUnorderedMatch({
+const evaluator = createTrajectoryMatchEvaluator({
+  trajectoryMatchMode: "unordered",
+});
+
+const result = await evaluator({
   outputs,
   referenceOutputs,
 });
@@ -229,26 +269,36 @@ console.log(result)
 }
 ```
 
-#### Subset and superset match
+`"unordered"` is useful is if you want to ensure that specific tools are called at some point in the trajectory, but you don't necessarily need them to be in message order (e.g. the agent called a company policy retrieval tool at an arbitrary point in an interaction before authorizing spend for a pizza party).
+
+**Note:** If you would like to configure the way this evaluator checks for tool call equality, see [this section](#tool-args-match-modes).
 
-There are other evaluators for checking partial trajectory matches (ensuring that a trajectory contains a subset and superset of tool calls compared to a reference trajectory).
+### Subset and superset match
+
+The `"subset"` and `"superset"` modes match partial trajectories (ensuring that a trajectory contains a subset/superset of tool calls contained in a reference trajectory).
 
 ```ts
-import { trajectorySubset } from "agentevals";
-// import { trajectorySuperset } from "agentevals";
+import { createTrajectoryMatchEvaluator } from "agentevals";
 
 const outputs = [
   { role: "user", content: "What is the weather in SF and London?" },
   {
     role: "assistant",
+    content: "",
     tool_calls: [{
       function: {
         name: "get_weather",
         arguments: JSON.stringify({ city: "SF and London" }),
       }
+    }, {
+      "function": {
+        name: "accuweather_forecast",
+        arguments: JSON.stringify({"city": "SF and London"}),
+      }
     }],
   },
   { role: "tool", content: "It's 80 degrees and sunny in SF, and 90 degrees and rainy in London." },
+  { role: "tool", content: "Unknown." },
   { role: "assistant", content: "The weather in SF is 80 degrees and sunny. In London, it's 90 degrees and rainy."},
 ];
 
@@ -256,27 +306,25 @@ const referenceOutputs = [
   { role: "user", content: "What is the weather in SF and London?" },
   {
     role: "assistant",
+    content: "",
     tool_calls: [
       {
         function: {
           name: "get_weather",
-          arguments: JSON.stringify({ city: "San Francisco" }),
-        }
-      },
-      {
-        function: {
-          name: "get_weather",
-          arguments: JSON.stringify({ city: "London" }),
+          arguments: JSON.stringify({ city: "SF and London" }),
         }
       },
     ],
   },
-  { role: "tool", content: "It's 80 degrees and sunny in San Francisco." },
-  { role: "tool", content: "It's 90 degrees and rainy in London." },
+  { role: "tool", content: "It's 80 degrees and sunny in San Francisco, and 90 degrees and rainy in London." },
   { role: "assistant", content: "The weather in SF is 80˚ and sunny. In London, it's 90˚ and rainy." },
 ];
 
-const result = await trajectorySubset({
+const evaluator = createTrajectoryMatchEvaluator({
+  trajectoryMatchMode: "superset", // or "subset"
+});
+
+const result = await evaluator({
   outputs,
   referenceOutputs,
 });
@@ -286,16 +334,145 @@ console.log(result)
 
 ```
 {
-    'key': 'trajectory_subset',
+    'key': 'trajectory_superset_match',
     'score': true,
 }
 ```
 
-#### Trajectory LLM-as-judge
+`"superset"` is useful if you want to ensure that some key tools were called at some point in the trajectory, but an agent calling extra tools is still acceptable. `"subset"` is the inverse and is useful if you want to ensure that the agent did not call any tools beyond the expected ones.
+
+**Note:** If you would like to configure the way this evaluator checks for tool call equality, see [this section](#tool-args-match-modes).
+
+### Tool args match modes
+
+When checking equality between tool calls, the above evaluators will require that all tool call arguments are the exact same by default. You can configure this behavior in the following ways:
+
+- Treating any two tool calls for the same tool as equivalent by setting `tool_args_match_mode="ignore"` (Python) or `toolArgsMatchMode: "ignore"` (TypeScript)
+- Treating a tool call as equivalent if it contain as subset/superset of args compared to a reference tool call of the same name with `tool_args_match_mode="subset"/"superset"` (Python) or `toolArgsMatchMode: "subset"/"superset` (TypeScript)
+- Setting custom matchers for all calls of a given tool using the `tool_args_match_overrides` (Python) or `toolArgsMatchOverrides` (TypeScript) param
+
+You can set both of these parameters at the same time. `tool_args_match_overrides` will take precendence over `tool_args_match_mode`.
+
+`tool_args_match_overrides`/`toolArgsMatchOverrides` takes a dictionary whose keys are tool names and whose values are either `"exact"`, `"ignore"`, a list of fields within the tool call that must match exactly, or a comparator function that takes two arguments and returns whether they are equal:
+
+```python
+ToolArgsMatchMode = Literal["exact", "ignore", "subset", "superset"]
+
+ToolArgsMatchOverrides = dict[str, Union[ToolArgsMatchMode, list[str],  Callable[[dict, dict], bool]]]
+```
+
+Here's an example that allows case insensitivity for the arguments to a tool named `get_weather`:
+
+```ts
+import { createTrajectoryMatchEvaluator } from "agentevals";
+
+const outputs = [
+  { role: "user", content: "What is the weather in SF?" },
+  {
+    role: "assistant",
+    content: "",
+    tool_calls: [{
+      function: {
+        name: "get_weather",
+        arguments: JSON.stringify({ city: "san francisco" })
+      },
+    }]
+  },
+  { role: "tool", content: "It's 80 degrees and sunny in SF." },
+  { role: "assistant", content: "The weather in SF is 80 degrees and sunny." },
+];
+
+const referenceOutputs = [
+  { role: "user", content: "What is the weather in San Francisco?" },
+  {
+    role: "assistant",
+    content: "",
+    tool_calls: [{
+      function: {
+        name: "get_weather",
+        arguments: JSON.stringify({ city: "San Francisco" })
+      }
+    }]
+  },
+  { role: "tool", content: "It's 80 degrees and sunny in San Francisco." },
+];
+
+const evaluator = createTrajectoryMatchEvaluator({
+  trajectoryMatchMode: "strict",
+  toolArgsMatchMode: "exact",  // Default value
+  toolArgsMatchOverrides: {
+    get_weather: (x, y) => {
+      return typeof x.city === "string" &&
+        typeof y.city === "string" &&
+        x.city.toLowerCase() === y.city.toLowerCase();
+    },
+  }
+});
+
+const result = await evaluator({
+  outputs,
+  referenceOutputs,
+});
+
+console.log(result);
+```
+
+```
+{
+  'key': 'trajectory_strict_match',
+  'score': true,
+}
+```
+
+This flexibility allows you to handle cases where you want looser equality for LLM generated arguments (`"san francisco"` to equal `"San Francisco"`) for only specific tool calls.
+
+## Trajectory LLM-as-judge
+
+The LLM-as-judge trajectory evaluator that uses an LLM to evaluate the trajectory. Unlike the trajectory match evaluators, it doesn't require a reference trajectory. Here's an example:
+
+```ts
+import {
+  createTrajectoryLLMAsJudge,
+  TRAJECTORY_ACCURACY_PROMPT,
+} from "agentevals";
+
+const evaluator = createTrajectoryLLMAsJudge({
+  prompt: TRAJECTORY_ACCURACY_PROMPT,
+  model: "openai:o3-mini",
+});
+
+const outputs = [
+  {role: "user", content: "What is the weather in SF?"},
+  {
+    role: "assistant",
+    content: "",
+    tool_calls: [
+      {
+        function: {
+          name: "get_weather",
+          arguments: JSON.stringify({ city: "SF" }),
+        }
+      }
+    ],
+  },
+  {role: "tool", content: "It's 80 degrees and sunny in SF."},
+  {role: "assistant", content: "The weather in SF is 80 degrees and sunny."},
+];
+
+const result = await evaluator({ outputs });
+
+console.log(result)
+```
+
+```
+{
+    'key': 'trajectory_accuracy',
+    'score': True,
+    'comment': 'The provided agent trajectory is reasonable...'
+}
+```
 
-The LLM-as-judge trajectory evaluator that uses an LLM to evaluate the trajectory. Unlike the other trajectory evaluators, it doesn't require a reference trajectory,
-and supports 
-This allows for more flexibility in the trajectory comparison:
+If you have a reference trajectory, you can add an extra variable to your prompt and pass in the reference trajectory. Below, we use the prebuilt  `TRAJECTORY_ACCURACY_PROMPT_WITH_REFERENCE` prompt, which contains a `reference_outputs` variable:
 
 ```ts
 import {
@@ -312,6 +489,7 @@ const outputs = [
   {role: "user", content: "What is the weather in SF?"},
   {
     role: "assistant",
+    content: "",
     tool_calls: [
       {
         function: {
@@ -328,6 +506,7 @@ const referenceOutputs = [
   {role: "user", content: "What is the weather in SF?"},
   {
     role: "assistant",
+    content: "",
     tool_calls: [
       {
         function: {
@@ -379,7 +558,7 @@ const fewShotExamples = [
 
 See the [`openevals`](https://github.com/langchain-ai/openevals?tab=readme-ov-file#llm-as-judge) repo for a fully up to date list of parameters.
 
-### Graph trajectory
+## Graph trajectory
 
 For frameworks like [LangGraph](https://github.com/langchain-ai/langgraph) that model agents as graphs, it can be more convenient to represent trajectories in terms of nodes visited rather than messages. `agentevals` includes a category of evaluators called **graph trajectory** evaluators that are designed to work with this format, as well as convenient utilities for extracting trajectories from a LangGraph thread, including different conversation turns and interrupts.
 
@@ -404,7 +583,7 @@ const evaluator: ({ inputs, outputs, referenceOutputs, ...extra }: {
 
 Where `inputs` is a list of inputs (or a dict with a key named `"inputs"`) to the graph whose items each represent the start of a new invocation in a thread, `results` representing the final output from each turn in the thread, and `steps` representing the internal steps taken for each turn.
 
-#### Graph trajectory LLM-as-judge
+### Graph trajectory LLM-as-judge
 
 This evaluator is similar to the `trajectory_llm_as_judge` evaluator, but it works with graph trajectories instead of message trajectories. Below, we set up a LangGraph agent, extract a trajectory from it using the built-in utils, and pass it to the evaluator. First, let's setup our graph, call it, and then extract the trajectory:
 
@@ -514,7 +693,7 @@ console.log(res);
 }
 ```
 
-Note that though this evaluator takes the typical `inputs`, `outputs`, and `referenceOutputs` parameters, it internally combines `inputs` and `outputs` to form a `thread`. Therefore, if you want to customize the prompt, your prompt should also contain a `thread` input variable:
+Note that though this evaluator takes the typical `inputs`, `outputs`, and `reference_outputs` parameters, it internally combines `inputs` and `outputs` to form a `thread`. Therefore, if you want to customize the prompt, your prompt should also contain a `thread` input variable:
 
 ```ts
 const CUSTOM_PROMPT = `You are an expert data labeler.
@@ -546,18 +725,18 @@ const graphTrajectoryEvaluator = createGraphTrajectoryLLMAsJudge({
   model: "openai:o3-mini",
 })
 res = await graphTrajectoryEvaluator(
-  inputs=extractedTrajectory.inputs,
-  outputs=extractedTrajectory.outputs,
+  inputs: extractedTrajectory.inputs,
+  outputs: extractedTrajectory.outputs,
 )
 ```
 
-In order to format them properly into the prompt, `referenceOutputs` should be passed in as a `GraphTrajectory` object like `outputs`.
+In order to format them properly into the prompt, `reference_outputs` should be passed in as a `GraphTrajectory` object like `outputs`.
 
-Also note that like other LLM-as-judge evaluators, you can pass extra kwargs into the evaluator to format them into the prompt.
+Also note that like other LLM-as-judge evaluators, you can pass extra params into the evaluator to format them into the prompt.
 
-#### Graph trajectory strict match
+### Graph trajectory strict match
 
-The `graphTrajectoryStrictMatch` evaluator is a simple evaluator that checks if the steps in the provided graph trajectory match the reference trajectory exactly.
+The `graph_trajectory_strict_match` evaluator is a simple evaluator that checks if the steps in the provided graph trajectory match the reference trajectory exactly.
 
 ```ts
 import { tool } from "@langchain/core/tools";
@@ -626,15 +805,46 @@ console.log(result);
   'score': True,
 }
 ```
-## LangSmith Integration
+
+# Python Async Support
+
+All `agentevals` evaluators support Python [asyncio](https://docs.python.org/3/library/asyncio.html). As a convention, evaluators that use a factory function will have `async` put immediately after `create_` in the function name (for example, `create_async_trajectory_llm_as_judge`), and evaluators used directly will end in `async` (e.g. `trajectory_strict_match_async`).
+
+Here's an example of how to use the `create_async_llm_as_judge` evaluator asynchronously:
+
+```python
+from agentevals.trajectory.llm import create_async_trajectory_llm_as_judge
+
+evaluator = create_async_llm_as_judge(
+    prompt="What is the weather in {inputs}?",
+)
+
+result = await evaluator(inputs="San Francisco")
+```
+
+If you are using the OpenAI client directly, remember to pass in `AsyncOpenAI` as the `judge` parameter:
+
+```python
+from openai import AsyncOpenAI
+
+evaluator = create_async_llm_as_judge(
+    prompt="What is the weather in {inputs}?",
+    judge=AsyncOpenAI(),
+    model="o3-mini",
+)
+
+result = await evaluator(inputs="San Francisco")
+```
+
+# LangSmith Integration
 
 For tracking experiments over time, you can log evaluator results to [LangSmith](https://smith.langchain.com/), a platform for building production-grade LLM applications that includes tracing, evaluation, and experimentation tools.
 
-LangSmith currently offers two ways to run evals. We'll give a quick example of how to run evals using both.
+LangSmith currently offers two ways to run evals: a [pytest](https://docs.smith.langchain.com/evaluation/how_to_guides/pytest) (Python) or [Vitest/Jest](https://docs.smith.langchain.com/evaluation/how_to_guides/vitest_jest) integration and the `evaluate` function. We'll give a quick example of how to run evals using both.
 
-### Pytest or Vitest/Jest
+## Pytest or Vitest/Jest
 
-First, follow [these instructions](https://docs.smith.langchain.com/evaluation/how_to_guides/vitest_jest) to set up LangSmith's Vitest/Jest runner,
+First, follow [these instructions](https://docs.smith.langchain.com/evaluation/how_to_guides/pytest) to set up LangSmith's pytest runner, or these to set up [Vitest or Jest](https://docs.smith.langchain.com/evaluation/how_to_guides/vitest_jest),
 setting appropriate environment variables:
 
 ```bash
@@ -642,7 +852,6 @@ export LANGSMITH_API_KEY="your_langsmith_api_key"
 export LANGSMITH_TRACING="true"
 ```
 
-
 Then, set up a file named `test_trajectory.eval.ts` with the following contents:
 
 ```ts
@@ -670,6 +879,7 @@ ls.describe("trajectory accuracy", () => {
         {"role": "user", "content": "What is the weather in SF?"},
         {
             "role": "assistant",
+            "content": "",
             "tool_calls": [
                 {
                     "function": {
@@ -688,6 +898,7 @@ ls.describe("trajectory accuracy", () => {
         {"role": "user", "content": "What is the weather in SF?"},
         {
             "role": "assistant",
+            "content": "",
             "tool_calls": [
                 {
                     "function": {
@@ -717,7 +928,6 @@ Now, run the eval with your runner of choice:
 vitest run test_trajectory.eval.ts
 ```
 
-
 Feedback from the prebuilt evaluator will be automatically logged in LangSmith as a table of results like this in your terminal:
 
 ![Terminal results](/static/img/pytest_output.png)
@@ -726,7 +936,7 @@ And you should also see the results in the experiment view in LangSmith:
 
 ![LangSmith results](/static/img/langsmith_results.png)
 
-### Evaluate
+## Evaluate
 
 Alternatively, you can [create a dataset in LangSmith](https://docs.smith.langchain.com/evaluation/concepts#dataset-curation) and use your created evaluators with LangSmith's [`evaluate`](https://docs.smith.langchain.com/evaluation#8-run-and-view-results) function:
 
@@ -741,20 +951,21 @@ const trajectoryEvaluator = createTrajectoryLLMAsJudge({
 
 await evaluate(
   (inputs) => [
-        {role: "user", content: "What is the weather in SF?"},
-        {
-            role: "assistant",
-            tool_calls: [
-                {
-                    function: {
-                        name: "get_weather",
-                        arguments: json.dumps({"city": "SF"}),
-                    }
-                }
-            ],
-        },
-        {role: "tool", content: "It's 80 degrees and sunny in SF."},
-        {role: "assistant", content: "The weather in SF is 80 degrees and sunny."},
+      {role: "user", content: "What is the weather in SF?"},
+      {
+          role: "assistant",
+          content: "",
+          tool_calls: [
+              {
+                  function: {
+                      name: "get_weather",
+                      arguments: json.dumps({"city": "SF"}),
+                  }
+              }
+          ],
+      },
+      {role: "tool", content: "It's 80 degrees and sunny in SF."},
+      {role: "assistant", content: "The weather in SF is 80 degrees and sunny."},
     ],
   {
     data: datasetName,
@@ -763,7 +974,7 @@ await evaluate(
 );
 ```
 
-## Thank you!
+# Thank you!
 
 We hope that `agentevals` helps make evaluating your LLM agents easier!
 

package/dist/index.cjs

@@ -14,7 +14,7 @@ var __exportStar = (this && this.__exportStar) || function(m, exports) {
     for (var p in m) if (p !== "default" && !Object.prototype.hasOwnProperty.call(exports, p)) __createBinding(exports, m, p);
 };
 Object.defineProperty(exports, "__esModule", { value: true });
-exports.GRAPH_TRAJECTORY_ACCURACY_PROMPT = exports.createGraphTrajectoryLLMAsJudge = exports.TRAJECTORY_ACCURACY_PROMPT_WITH_REFERENCE = exports.TRAJECTORY_ACCURACY_PROMPT = exports.createTrajectoryLLMAsJudge = exports.trajectoryUnorderedMatch = exports.trajectorySuperset = exports.trajectorySubset = exports.trajectoryStrictMatch = void 0;
+exports.GRAPH_TRAJECTORY_ACCURACY_PROMPT = exports.createGraphTrajectoryLLMAsJudge = exports.TRAJECTORY_ACCURACY_PROMPT_WITH_REFERENCE = exports.TRAJECTORY_ACCURACY_PROMPT = exports.createTrajectoryLLMAsJudge = exports.createTrajectoryMatchEvaluator = exports.trajectoryUnorderedMatch = exports.trajectorySuperset = exports.trajectorySubset = exports.trajectoryStrictMatch = void 0;
 var strict_js_1 = require("./trajectory/strict.cjs");
 Object.defineProperty(exports, "trajectoryStrictMatch", { enumerable: true, get: function () { return strict_js_1.trajectoryStrictMatch; } });
 var subset_js_1 = require("./trajectory/subset.cjs");
@@ -23,6 +23,8 @@ var superset_js_1 = require("./trajectory/superset.cjs");
 Object.defineProperty(exports, "trajectorySuperset", { enumerable: true, get: function () { return superset_js_1.trajectorySuperset; } });
 var unordered_js_1 = require("./trajectory/unordered.cjs");
 Object.defineProperty(exports, "trajectoryUnorderedMatch", { enumerable: true, get: function () { return unordered_js_1.trajectoryUnorderedMatch; } });
+var match_js_1 = require("./trajectory/match.cjs");
+Object.defineProperty(exports, "createTrajectoryMatchEvaluator", { enumerable: true, get: function () { return match_js_1.createTrajectoryMatchEvaluator; } });
 var llm_js_1 = require("./trajectory/llm.cjs");
 Object.defineProperty(exports, "createTrajectoryLLMAsJudge", { enumerable: true, get: function () { return llm_js_1.createTrajectoryLLMAsJudge; } });
 Object.defineProperty(exports, "TRAJECTORY_ACCURACY_PROMPT", { enumerable: true, get: function () { return llm_js_1.TRAJECTORY_ACCURACY_PROMPT; } });

package/dist/index.d.ts

@@ -2,6 +2,7 @@ export { trajectoryStrictMatch } from "./trajectory/strict.js";
 export { trajectorySubset } from "./trajectory/subset.js";
 export { trajectorySuperset } from "./trajectory/superset.js";
 export { trajectoryUnorderedMatch } from "./trajectory/unordered.js";
+export { createTrajectoryMatchEvaluator, type TrajectoryMatchMode, } from "./trajectory/match.js";
 export { createTrajectoryLLMAsJudge, TRAJECTORY_ACCURACY_PROMPT, TRAJECTORY_ACCURACY_PROMPT_WITH_REFERENCE, } from "./trajectory/llm.js";
 export { createGraphTrajectoryLLMAsJudge, GRAPH_TRAJECTORY_ACCURACY_PROMPT, } from "./graph_trajectory/llm.js";
 export * from "./types.js";

package/dist/index.js

@@ -2,6 +2,7 @@ export { trajectoryStrictMatch } from "./trajectory/strict.js";
 export { trajectorySubset } from "./trajectory/subset.js";
 export { trajectorySuperset } from "./trajectory/superset.js";
 export { trajectoryUnorderedMatch } from "./trajectory/unordered.js";
+export { createTrajectoryMatchEvaluator, } from "./trajectory/match.js";
 export { createTrajectoryLLMAsJudge, TRAJECTORY_ACCURACY_PROMPT, TRAJECTORY_ACCURACY_PROMPT_WITH_REFERENCE, } from "./trajectory/llm.js";
 export { createGraphTrajectoryLLMAsJudge, GRAPH_TRAJECTORY_ACCURACY_PROMPT, } from "./graph_trajectory/llm.js";
 export * from "./types.js";