Add Amazon bedrock doc (#116)

nicoloboschi · mendonk · web-flow · commit e3d5ae572f82 · 2023-10-19T10:23:21.000-04:00
Co-authored-by: Mendon Kissling &lt;59585235+mendonk@users.noreply.github.com&gt;
diff --git a/SUMMARY.md b/SUMMARY.md
@@ -58,6 +58,7 @@
   * [OpenAI](configuration-resources/large-language-models-llms/open-ai-configuration.md)
   * [Hugging Face](configuration-resources/large-language-models-llms/hugging-face-configuration.md)
   * [Google Vertex AI](configuration-resources/large-language-models-llms/vertex-configuration.md)
+  * [Amazon Bedrock](configuration-resources/large-language-models-llms/bedrock-configuration.md)
 * [Data storage](configuration-resources/data-storage/README.md)
   * [Astra](configuration-resources/data-storage/astra.md)
   * [Cassandra](configuration-resources/data-storage/cassandra.md)
diff --git a/building-applications/build-a-sample-app.md b/building-applications/build-a-sample-app.md
@@ -98,7 +98,7 @@ pipeline:
       completion-field: "value"
       messages:
         - role: user
-          content: "What can you tell me about {{% value}} ?"
+          content: "What can you tell me about {{ value}} ?"
 ```
 
 {% hint style="info" %}
@@ -196,7 +196,7 @@ application:
         configuration:
           completion-field: "value"
           messages:
-          - content: "What can you tell me about {{% value}} ?"
+          - content: "What can you tell me about {{ value}} ?"
             role: "user"
           model: "gpt-35-turbo"
         resources:
diff --git a/building-applications/error-handling.md b/building-applications/error-handling.md
@@ -26,7 +26,7 @@ pipeline:
     configuration:
       model: "text-embedding-ada-002"
       embeddings-field: "value.question_embeddings"
-      text: "{{% value.question }}"
+      text: "{{ value.question }}"
 ```
 
 Source agents do not have configurable error handling.
diff --git a/building-applications/pipelines.md b/building-applications/pipelines.md
@@ -21,7 +21,7 @@ pipeline:
       completion-field: "value"
       messages:
         - role: user
-          content: "What can you tell me about {{% value}} ?"
+          content: "What can you tell me about {{ value}} ?"
 ```
 
 ## Pipeline configuration
diff --git a/configuration-resources/large-language-models-llms/README.md b/configuration-resources/large-language-models-llms/README.md
@@ -14,4 +14,7 @@ layout:
 
 # Large Language Models (LLMs)
 
-<table data-view="cards"><thead><tr><th></th><th data-hidden></th><th data-hidden></th><th data-hidden data-card-target data-type="content-ref"></th></tr></thead><tbody><tr><td>open-ai-configuration</td><td></td><td></td><td><a href="open-ai-configuration.md">open-ai-configuration.md</a></td></tr><tr><td>hugging-face-configuration</td><td></td><td></td><td><a href="hugging-face-configuration.md">hugging-face-configuration.md</a></td></tr><tr><td>vertex-configuration</td><td></td><td></td><td><a href="vertex-configuration.md">vertex-configuration.md</a></td></tr></tbody></table>
+* [OpenAI](open-ai-configuration.md)
+* [Hugging Face](hugging-face-configuration.md)
+* [Google Vertex AI](vertex-configuration.md)
+* [Amazon Bedrock](bedrock-configuration.md)
diff --git a/configuration-resources/large-language-models-llms/bedrock-configuration.md b/configuration-resources/large-language-models-llms/bedrock-configuration.md
@@ -0,0 +1,30 @@
+# open-ai-configuration
+
+The resource `bedrock-configuration` configures access to the Amazon Bedrock models for embeddings and completions.
+
+`configuration.yaml`
+```yaml
+configuration:
+  resources:
+    - type: "bedrock-configuration"
+      name: "Bedrock"
+      configuration:
+        access-key: "${secrets.bedrock.access-key}"
+        secret-key: "${secrets.bedrock.secret-key}"
+        region: "${secrets.bedrock.region}"
+```
+
+`secrets.yaml`
+```yaml
+secrets:
+  - id: bedrock
+    data:
+      access-key: "${BEDROCK_ACCESS_KEY}"
+      secret-key: "${BEDROCK_SECRET_KEY}"
+      region: "${BEDROCK_REGION}"
+```
+
+
+## Configuration
+
+Check out the full configuration properties in the [API Reference page](../../building-applications/api-reference/resources.md#bedrock-configuration).
diff --git a/configuration-resources/large-language-models-llms/hugging-face-configuration.md b/configuration-resources/large-language-models-llms/hugging-face-configuration.md
@@ -20,6 +20,6 @@ secrets:
     data:
       access-key: "${HF_ACCESS_KEY}"
 ```
-### API Reference
+## Configuration
 
-<table><thead><tr><th width="158.33333333333331">Label</th><th width="139">Type</th><th>Description</th></tr></thead><tbody><tr><td>access-key</td><td><br></td><td><p>Your hugging face token. </p></td></tr><tr><td>inference-url</td><td><br></td><td><p>Hugging face inference url.</p><p></p><p>Default value is <code>https://api-inference.huggingface.co</code></p></td></tr></tbody></table>
+Check out the full configuration properties in the [API Reference page](../../building-applications/api-reference/resources.md#hugging-face-configuration).
diff --git a/configuration-resources/large-language-models-llms/open-ai-configuration.md b/configuration-resources/large-language-models-llms/open-ai-configuration.md
@@ -45,6 +45,6 @@ secrets:
       access-key: "${AZURE_ACCESS_KEY}"
 ```
 
+## Configuration
 
-## API reference
-<table><thead><tr><th width="158.33333333333331">Labele</th><th width="111">Type</th><th>Description</th></tr></thead><tbody><tr><td>url</td><td>string</td><td><p><b>Only for Azure OpenAI provider</b>Base url for the OpenAI api. </p><p></p><p>Example: https://COMPANY.openai.azure.com</p></td></tr><tr><td>access-key</td><td>string</td><td><p>Your OpenAI key</p><p></p></td></tr><tr><td>provider</td><td>string</td><td><p>OpenAI provider type. Supported values are:</p><ul><li>“openai” (default)</li><li>“azure”</li></ul></td></tr></tbody></table>
+Check out the full configuration properties in the [API Reference page](../../building-applications/api-reference/resources.md#open-ai-configuration).
diff --git a/configuration-resources/large-language-models-llms/vertex-configuration.md b/configuration-resources/large-language-models-llms/vertex-configuration.md
@@ -33,6 +33,6 @@ secrets:
 For authenticating it's suggested to use the service account JSON since tokens are normally generated with a short TTL.
 
 
-## API Reference
+## Configuration
 
-<table><thead><tr><th width="229.33333333333331">Label</th><th width="110">Type</th><th>Description</th></tr></thead><tbody><tr><td>url</td><td><br>String</td><td><p>URL connection for the Vertex API.</p></td></tr><tr><td>serviceAccountJson</td><td><br>String</td><td><p>Specify a custom service account. Refer to the <a href="https://cloud.google.com/iam/docs/keys-create-delete">GCP documentation</a> on how to download it</p></td></tr><tr><td>token</td><td><br>String</td><td><p>Access key for the Vertex API. You can generate a short-life token with <code>gcloud auth print-access-token</code>.</p></td></tr><tr><td>region</td><td>String<br></td><td><p>GCP region for the Vertex API. </p></td></tr><tr><td>project</td><td>String<br></td><td>GCP project name for the Vertex API.</td></tr></tbody></table>
+Check out the full configuration properties in the [API Reference page](../../building-applications/api-reference/resources.md#vertex-configuration).
diff --git a/patterns/rag-pattern.md b/patterns/rag-pattern.md
@@ -36,7 +36,7 @@ pipeline:
     configuration:
       model: "${secrets.open-ai.embeddings-model}" # This needs to match the name of the model deployment, not the base model
       embeddings-field: "value.question_embeddings"
-      text: "{{% value.question }}"
+      text: "{{ value.question }}"
       flush-interval: 0
   - name: "lookup-related-documents"
     type: "query-vector-db"
@@ -90,11 +90,11 @@ pipeline:
               Do not provide information that is not related to the LangStream project.
             
               Documents:
-              {{%# value.related_documents}}
-              {{% text}}
-              {{%/ value.related_documents}}
+              {{# value.related_documents}}
+              {{ text}}
+              {{/ value.related_documents}}
         - role: user
-          content: "{{% value.question}}"          
+          content: "{{ value.question}}"          
 ```
 
 ### What’s next?
diff --git a/pipeline-agents/agent-messaging.md b/pipeline-agents/agent-messaging.md
@@ -25,7 +25,7 @@ If the input message to an agent is in plain text, then you refer to it as “\{
 ```yaml
 messages:
   - role: user
-    content: "What can you tell me about {{% value}} ?"
+    content: "What can you tell me about {{ value}} ?"
 ```
 
 Or in the query agent if you wanted to include the input message as a part of the query you would:
diff --git a/pipeline-agents/ai-actions/README.md b/pipeline-agents/ai-actions/README.md
@@ -14,6 +14,7 @@ layout:
 
 # AI Actions
 
-<table data-view="cards"><thead><tr><th></th><th data-hidden></th><th data-hidden></th><th data-hidden data-card-target data-type="content-ref"></th></tr></thead><tbody><tr><td>ai-chat-completions</td><td>ai-chat-completions</td><td></td><td><a href="ai-chat-completions.md">ai-chat-completions.md</a></td></tr><tr><td>ai-text-completions</td><td>ai-text-completions</td><td></td><td><a href="ai-text-completions.md">ai-text-completions.md</a></td></tr><tr><td>compute-ai-embeddings</td><td></td><td></td><td><a href="compute-ai-embeddings.md">compute-ai-embeddings.md</a></td></tr>
-<tr><td>flare-controller</td><td></td><td></td><td><a href="flare-controller.md">flare-controller.md</a></td></tr></tbody></table>
-
+* [Chat completions](ai-chat-completions.md)
+* [Text completions](ai-text-completions.md)
+* [Embeddings](compute-ai-embeddings.md)
+* [FLARE](flare-controller.md)
diff --git a/pipeline-agents/ai-actions/ai-chat-completions.md b/pipeline-agents/ai-actions/ai-chat-completions.md
@@ -19,15 +19,18 @@ Given the AI model specified in an application's configuration resources, this a
 
 ## Using OpenAI chat models
 
-The `ai-chat-completions` for OpenAI uses the /v1/chat/completions endpoint. Refer to the [OpenAI documentation](https://platform.openai.com/docs/models/model-endpoint-compatibility) to know which models are compatible.
+> The `ai-chat-completions` for OpenAI uses the /v1/chat/completions endpoint. Refer to the [OpenAI documentation](https://platform.openai.com/docs/models/model-endpoint-compatibility) to know which models are compatible.
+
+Setup the OpenAI LLM [configuration](../../configuration-resources/large-language-models-llms/open-ai-configuration.md).
+Add the `ai-chat-completions` agent:
 
 ```yaml
 pipeline:
   - name: "ai-chat-completions"
     type: "ai-chat-completions"
     output: "history-topic"
     configuration:
-      model: "${secrets.open-ai.chat-completions-model}" # This needs to be set to the model deployment name, not the base name
+      model: "gpt-3.5-turbo"
       # on the log-topic we add a field with the answer
       completion-field: "value.answer"
       # we are also logging the prompt we sent to the LLM
@@ -45,26 +48,86 @@ pipeline:
       min-chunks-per-message: 10
       messages:
         - role: user
-          content: "You are a helpful assistant. Below you can find a question from the user. Please try to help them the best way you can.\n\n{{ value.question}}"
+          content: "You are a helpful assistant. Below you can find a question from the user. Please try to help them the best way you can.\n\n{{ value.question }}"
 ```
 
 ## Using VertexAI chat models
 
-Refer to the [VertexAI documentation](https://cloud.google.com/vertex-ai/docs/generative-ai/model-reference/text-chat) to know which models are compatible.
+> Refer to the [VertexAI documentation](https://cloud.google.com/vertex-ai/docs/generative-ai/model-reference/text-chat) to know which models are compatible.
+
+Setup the Vertex LLM [configuration](../../configuration-resources/large-language-models-llms/vertex-configuration.md).
+Add the `ai-chat-completions` agent:
 
 ```yaml
 pipeline:
   - name: "ai-chat-completions"
     type: "ai-chat-completions"
     configuration:
-      model: "bert-base-uncased"
+      model: "chat-bison"
+      max-tokens: 100
       completion-field: "value.chatresult"
       log-field: "value.request"
       messages:
         - role: user
-          content: "{{ value }} [MASK]"
+          content: "You are a helpful assistant. Below you can find a question from the user. Please try to help them the best way you can.\n\n{{ value.question }}"
+```
+
+
+## Using Amazon Bedrock AI21 Jurassic-2 models
+
+> Refer to the [Amazon documentation](https://docs.aws.amazon.com/bedrock/latest/userguide/model-parameters.html#model-parameters-jurassic2) for other parameters and options.
+
+Set up the Amazon Bedrock LLM [configuration](../../configuration-resources/large-language-models-llms/bedrock-configuration.md).
+Add the `ai-chat-completions` agent:
+
+```yaml
+pipeline:
+  - name: "ai-chat-completions"
+    type: "ai-chat-completions"
+    configuration:
+      model: "ai21.j2-mid-v1"
+      completion-field: "value.answer"
+      options:
+        request-parameters:
+          # here you can add all the supported parameters
+          temperature: 0.5
+          maxTokens: 300
+        # expression to retrieve the completion from the response JSON. It varies depending on the model 
+        response-completions-expression: "completions[0].data.text"
+      messages:
+        - content: "{{ value.question }}"
+```
+
+
+
+## Using Amazon Bedrock Anthropic Claude models
+
+> Refer to the [Amazon documentation](https://docs.aws.amazon.com/bedrock/latest/userguide/model-parameters.html#model-parameters-claude) to learn other parameters and options.
+
+Setup the Amazon Bedrock LLM [configuration](../../configuration-resources/large-language-models-llms/bedrock-configuration.md).
+Add the `ai-chat-completions` agent:
+
+```yaml
+pipeline:
+  - name: "ai-chat-completions"
+    type: "ai-chat-completions"
+    configuration:
+      model: "anthropic.claude-v2"
+      completion-field: "value.answer"
+      options:
+        request-parameters:
+          # here you can add all the supported parameters
+          temperature: 0.5
+          max_tokens_to_sample: 300
+          top_p: 0.9
+          top_k: 250
+        # expression to retrieve the completion from the response JSON. It varies depending on the model 
+        response-completions-expression: "completion"
+      messages:
+        - content: "{{ value.question }}"
 ```
 
+
 ### Prompt limitations
 
 Most public LLMs have character limits on message content. It is up to the application developer to ensure the combination of preset prompt text and an input message stays under that limit.
@@ -88,8 +151,4 @@ Some public LLMs offer a free tier and then automatically begin charging per pro
 
 ### Configuration
 
-<table><thead><tr><th width="182.33333333333331">Label</th><th width="162">Type</th><th>Description</th></tr></thead><tbody><tr><td>model</td><td>string (required)</td><td><p>Given the AI model set in the configuration resources, this is the corresponding model name to use.<br></p><p>Example using the OpenAI model: “gpt-3.5-turbo”</p></td></tr><tr><td>completion-field</td><td>string (required)</td><td><p>The name of an additional field that will be added to the output message data containing the LLM completion.<br></p><p>Provide in the form: “value.&#x3C;field-name>” (do not include mustache brackets, this not a templated value).</p></td></tr><tr><td>log-field</td><td>string (optional)</td><td>This is the final prompt that was submitted to the model.</td></tr><tr><td>stream-to-topic</td><td>string (optional)</td><td></td></tr><tr><td>stream-response-completion-field</td><td>string (optional)</td><td></td></tr><tr><td>messages</td><td>object[]</td><td><p>A collection of LLM prompt messages, see the reference table.</p><p></p><p>Example collection:</p><ul><li>role: user<br>message: “answer the question: {{% value }}”</li><li>role: system<br>message: “You are a friendly customer service agent”</li></ul></td></tr></tbody></table>
-
-#### messages
-
-<table><thead><tr><th width="124.33333333333331">Label</th><th width="165">Type</th><th>Description</th></tr></thead><tbody><tr><td>role</td><td>string (required)</td><td>Values are limited to what the chosen LLM model supports. Typical values are “user” or “system”.</td></tr><tr><td>content</td><td>string (required)</td><td><p>A templated value that will be sent to the LLM. Refer to message structure and templating for templating options.<br><br></p><p>Example of combining preset prompt and entire input value:</p><p>“You are a friendly customer service agent and like answering questions like: {{% value }}”</p><p><br></p><p>Example of combining preset prompt and specific input value:</p><p>“You are a friendly customer service agent speaking with {{% value.first-name}}. Help them with the question: {{% value.user-question }}”</p></td></tr></tbody></table>
+Check out the full configuration properties in the [API Reference page](../../building-applications/api-reference/agents.md#ai-chat-completions).
diff --git a/pipeline-agents/ai-actions/ai-text-completions.md b/pipeline-agents/ai-actions/ai-text-completions.md
diff --git a/pipeline-agents/ai-actions/compute-ai-embeddings.md b/pipeline-agents/ai-actions/compute-ai-embeddings.md
diff --git a/pipeline-agents/input-and-output/webcrawler-source.md b/pipeline-agents/input-and-output/webcrawler-source.md
diff --git a/pipeline-agents/text-processors/rerank.md b/pipeline-agents/text-processors/rerank.md
