# Moving from `trulens-eval`

This document highlights the changes required to migrate from `trulens-eval` (any version) to `trulens` v1.0+.

The biggest change is that the `trulens` library now consists of several interoperable modules, each of which can be installed and used independently. This allows users to mix and match components to suit their needs without needing to install the entire library.

When running `pip install trulens`, the following base modules are installed:

- `trulens-core`: core module that provides the main functionality for TruLens.  
- `trulens-feedback`: The module that provides LLM-based evaluation and feedback function definitions.  
- `trulens-dashboard`: The module that supports the streamlit dashboard and evaluation visualizations.

Furthermore, the following additional modules can be installed separately:  
- `trulens-benchmark`: provides benchmarking functionality for evaluating feedback functions on your dataset.

Instrumentation libraries used to instrument specific frameworks like LangChain and LlamaIndex are now packaged separately and imported under the `trulens.apps` namespace. For example, to use TruChain to instrument a LangChain app, run `pip install trulens-apps-langchain` and import it as follows:

```
from trulens.apps.langchain import TruChain
```

Similarly, providers are now packaged separately from the core library. To use a specific provider, install the corresponding package and import it as follows:

```
from trulens.providers.openai import OpenAI
```

To find a full list of providers, please refer to the [API Reference](/content/reference/index.html).

## Common Import Changes

As a result of these changes, the package structure for TruLens differs from TruLens-Eval. Here are some common import changes you may need to make:

| TruLens Eval | TruLens | Additional Dependencies |
| --- | --- | --- |
| `trulens_eval.Tru` | [trulens.core.TruSession](/content/reference/trulens/core/#trulens.core.TruSession "            TruSession"/index.html) |  |
| `trulens_eval.Feedback` | [trulens.core.Feedback](/content/reference/trulens/core/#trulens.core.Feedback "            Feedback"/index.html) |  |
| `trulens_eval.Select` | [trulens.core.Select](/content/reference/trulens/core/#trulens.core.Select "            Select"/index.html) |  |
| `trulens_eval.TruCustomApp`, `TruSession().Custom(...)` | [trulens.apps.app.TruApp](/content/reference/trulens/apps/app/#trulens.apps.app.TruApp "            TruApp"/index.html) |  |
| `trulens_eval.TruChain`, `Tru().Chain(...)` | [`TruSession().App(...)`](/content/reference/trulens/core/session/#trulens.core.session.TruSession.App "            App"/index.html) or [trulens.apps.langchain.TruChain](/content/reference/trulens/apps/langchain/#trulens.apps.langchain.TruChain "            TruChain"/index.html) | `trulens-apps-langchain` |
| `trulens_eval.TruLlama`, `Tru().Llama(...)` | [`TruSession().App(...)`](/content/reference/trulens/core/session/#trulens.core.session.TruSession.App "            App"/index.html) or [trulens.apps.llamaindex.TruLlama](/content/reference/trulens/apps/llamaindex/#trulens.apps.llamaindex.TruLlama "            TruLlama"/index.html) | `trulens-apps-llamaindex` |
| `trulens_eval.TruRails`, `Tru().Rails(...)` | [`TruSession().App(...)`](/content/reference/trulens/core/session/#trulens.core.session.TruSession.App "            App"/index.html) or [trulens.apps.nemo.TruRails](/content/reference/trulens/apps/nemo/#trulens.apps.nemo.TruRails "            TruRails"/index.html) | `trulens-apps-nemo` |
| `trulens_eval.OpenAI` | [trulens.providers.openai.OpenAI](/content/reference/trulens/providers/openai/#trulens.providers.openai.OpenAI "            OpenAI"/index.html) | `trulens-providers-openai` |
| `trulens_eval.Huggingface` | [trulens.providers.huggingface.Huggingface](/content/reference/trulens/providers/huggingface/#trulens.providers.huggingface.Huggingface "            Huggingface"/index.html) | `trulens-providers-huggingface` |
| `trulens_eval.guardrails.llama` | [trulens.apps.llamaindex.guardrails](/content/reference/trulens/apps/llamaindex/guardrails/#trulens.apps.llamaindex.guardrails "            trulens.apps.llamaindex.guardrails"/index.html) | `trulens-apps-llamaindex` |
| `Tru().run_dashboard()` | [`trulens.dashboard.run_dashboard()`](/content/reference/trulens/dashboard/#trulens.dashboard.run_dashboard "            run_dashboard"/index.html) | `trulens-dashboard` |
|  |  |  |

For Snowflake-based workflows, use the AI Observability **Evaluations** page in Snowsight instead of launching a local dashboard.

To find a specific definition, use the search functionality or go directly to the [API Reference](/content/reference/index.html).

## Automatic Migration with Grit

To assist you in migrating your codebase to _TruLens_ v1.0, we've published a `grit` pattern. You can migrate your codebase by using `grit` on the command line. See the TruLens migration grit pattern in the [stdlib repo](https://github.com/honeycombio/stdlib/blob/main/.grit/patterns/python/trulens_eval_migration.md).

To use Grit on the command line, follow these instructions:

### Install `grit`

You can install the Grit CLI from NPM:  
```
npm install --location=global @getgrit/cli
```  
Alternatively, you can install Grit with an installation script:  
```
curl -fsSL https://docs.grit.io/install | bash
```

### Apply automatic changes  
```
grit apply trulens_eval_migration
```  
Review and audit all changes carefully before committing. We recommend ensuring you have a clean working tree and recent backup before running the migration.