messages listlengths 2 46 |
|---|
[
{
"role": "user",
"content": "How does the response‑quality‑difference threshold influence when the router switches from the fallback model to a higher‑quality model?"
},
{
"role": "assistant",
"content": "The threshold defines the minimum predicted improvement a higher‑quality model must have o... |
[
{
"role": "user",
"content": "What’s the difference between default prompt routers and configured prompt routers, and when should I use each?"
},
{
"role": "assistant",
"content": "Default routers are pre‑configured by Amazon Bedrock and work out‑of‑the‑box with a limited set of Anthropic and Me... |
[
{
"role": "user",
"content": "How can I determine whether a specific model (e.g., Claude 3.5 Sonnet) is available in my AWS region?"
},
{
"role": "assistant",
"content": "Check the \"Regions supporting foundation model\" and \"Regions supporting inference profile\" columns in the supported model... |
[
{
"role": "user",
"content": "Can I configure a single prompt router to route between an Anthropic model and a Meta Llama model?"
},
{
"role": "assistant",
"content": "No. Intelligent prompt routing only routes between models **within the same model family**. You must create separate routers for... |
[
{
"role": "user",
"content": "How does intelligent prompt routing help reduce my Bedrock API costs while maintaining response quality?"
},
{
"role": "assistant",
"content": "The router predicts the response quality for each eligible model on a per‑request basis. It then selects the model that of... |
[
{
"role": "user",
"content": "What happens if I send a non‑English prompt to a router that’s optimized for English?"
},
{
"role": "assistant",
"content": "The routing predictions are trained on English data, so the quality estimates for non‑English prompts may be inaccurate. This can lead to sub... |
[
{
"role": "user",
"content": "How do I enable intelligent prompt routing using the AWS CLI and what metrics should I monitor to ensure it’s working as expected?"
},
{
"role": "assistant",
"content": "1️⃣ Enable routing with the CLI by creating a prompt‑router resource (e.g., `aws bedrock create-... |
[
{
"role": "user",
"content": "If Amazon releases a new Claude model next month, will I need to modify my existing prompt router to take advantage of it?"
},
{
"role": "assistant",
"content": "No. The routing service is designed to be future‑proof: it automatically incorporates newly available mo... |
[
{
"role": "user",
"content": "Can I store binary vector embeddings in Amazon S3 Vectors or Aurora PostgreSQL, or do I need to use a specific vector store?"
},
{
"role": "assistant",
"content": "No. Binary vector embeddings are supported only by Amazon OpenSearch Serverless and Amazon OpenSearch ... |
[
{
"role": "user",
"content": "How do I determine the correct number of dimensions to configure for my vector index across different vector stores?"
},
{
"role": "assistant",
"content": "Select the dimensions that match the embedding model you plan to use. The supported dimensions are:\n- Titan G... |
[
{
"role": "user",
"content": "What are the metadata size limits for Amazon S3 Vectors, and how should I store large text chunks if they exceed the filterable metadata quota?"
},
{
"role": "assistant",
"content": "Each vector in an S3 Vector index can have up to 40 KB of total metadata, of which ... |
[
{
"role": "user",
"content": "Why am I seeing data ingestion failures in an OpenSearch Managed cluster, and how can I resolve them?"
},
{
"role": "assistant",
"content": "Ingestion failures usually indicate that the domain does not have enough capacity. Increase the domain's provisioned IOPS and... |
[
{
"role": "user",
"content": "What is the curl command payload to create a FAISS vector index with binary embeddings in an OpenSearch Managed domain?"
},
{
"role": "assistant",
"content": "Use a PUT request with the following JSON body (replace placeholders with your values):\n```json\nPUT /<ind... |
[
{
"role": "user",
"content": "How can I enable efficient filtering on custom JSONB metadata in an Aurora PostgreSQL knowledge base?"
},
{
"role": "assistant",
"content": "Create a GIN index on the `custom_metadata` column to allow key‑value pair searches, and for frequently used numeric filters ... |
[
{
"role": "user",
"content": "What steps are needed to let Amazon Bedrock decrypt a KMS‑encrypted Secrets Manager secret that contains Pinecone credentials?"
},
{
"role": "assistant",
"content": "1. Store the Pinecone API key in a Secrets Manager secret (key name `apiKey`). 2. (Optional) Encrypt... |
[
{
"role": "user",
"content": "Do I need to adjust the vector dimension setting in Neptune Analytics if I change the Bedrock embedding model after the graph is created?"
},
{
"role": "assistant",
"content": "Yes. The vector dimension configured in the Neptune graph must always match the dimension... |
[
{
"role": "user",
"content": "How can I connect Amazon Bedrock to a MongoDB Atlas cluster over a private network instead of the public internet?"
},
{
"role": "assistant",
"content": "Create an AWS PrivateLink endpoint for your MongoDB Atlas VPC peering connection and configure the Atlas cluster... |
[
{
"role": "user",
"content": "Can I upload a UTF‑8 .txt file that contains Unicode characters, or does the plain‑text format only accept ASCII?"
},
{
"role": "assistant",
"content": "The plain‑text format for .txt files is limited to ASCII characters only. If your .txt file contains Unicode char... |
[
{
"role": "user",
"content": "What should I do if my PDF document is larger than 50 MB, which exceeds the size limit?"
},
{
"role": "assistant",
"content": "Split the PDF into smaller parts so that each resulting file is ≤ 50 MB, or convert the content to another supported format (e.g., Markdown... |
[
{
"role": "user",
"content": "If a JPEG or PNG image exceeds the 3.75 MB limit, will the ingestion succeed or fail?"
},
{
"role": "assistant",
"content": "The ingestion will fail for that image. The documentation states that JPEG and PNG files must not exceed 3.75 MB. You must resize or compress... |
[
{
"role": "user",
"content": "Can a single Amazon Bedrock knowledge base contain a mixture of .md, .pdf, .docx, and Excel files, or must each format be loaded into its own knowledge base?"
},
{
"role": "assistant",
"content": "Yes, a single knowledge base can ingest a heterogeneous collection of... |
[
{
"role": "user",
"content": "Do the 50 MB file‑size limits apply equally to files stored in Amazon S3 and to files coming from a custom data source?"
},
{
"role": "assistant",
"content": "Yes. The 50 MB size limit is a universal rule for all supported document formats, regardless of whether the... |
[
{
"role": "user",
"content": "How can I include tabular data from an Excel spreadsheet in my knowledge base, and are there any special considerations?"
},
{
"role": "assistant",
"content": "Upload the spreadsheet as a .xls or .xlsx file (both are supported). The ingestion process will extract th... |
[
{
"role": "user",
"content": "Is it possible to use JPEG or PNG images when my knowledge base source is an Amazon DynamoDB table?"
},
{
"role": "assistant",
"content": "No. Multimodal data such as JPEG and PNG files are allowed only when the source is Amazon S3 or a custom data source. Other sou... |
[
{
"role": "user",
"content": "How does Amazon Bedrock Guardrails affect inference latency, and what mechanisms are in place to minimize any added delay?"
},
{
"role": "assistant",
"content": "Guardrails evaluate each configured policy in parallel for the input, which reduces the time added compa... |
[
{
"role": "user",
"content": "What are the cost differences between a guardrail that blocks the input prompt versus one that only blocks the model response?"
},
{
"role": "assistant",
"content": "If the guardrail blocks the input prompt, you are charged only for the policy evaluation; the founda... |
[
{
"role": "user",
"content": "Can a single Guardrail be applied to both text and image foundation models, and does it require separate configurations for each modality?"
},
{
"role": "assistant",
"content": "Yes. A Guardrail can be referenced during inference for any text or image foundation mod... |
[
{
"role": "user",
"content": "True or false: Amazon Bedrock Guardrails can be applied to inference calls made by Bedrock Agents and Bedrock Knowledge Bases."
},
{
"role": "assistant",
"content": "True. Guardrails are supported for both Amazon Bedrock Agents and Amazon Bedrock Knowledge Bases, al... |
[
{
"role": "user",
"content": "How can I customize what the user sees when a guardrail blocks a model response—can I choose between a blocked message and masking the offending content?"
},
{
"role": "assistant",
"content": "Each policy in a Guardrail can be configured with either a *pre‑configure... |
[
{
"role": "user",
"content": "Is there a limit to the number of Guardrails an AWS account can create?"
},
{
"role": "assistant",
"content": "The documentation states that an account can have multiple Guardrails, each with its own configuration, but it does not specify a hard limit. In practice, ... |
[
{
"role": "user",
"content": "How do I reference a Guardrail when invoking a model using the InvokeModel API?"
},
{
"role": "assistant",
"content": "When calling InvokeModel (or InvokeModelWithResponseStream, Converse, ConverseStream), include the Guardrail identifier in the request payload unde... |
[
{
"role": "user",
"content": "Which AWS Marketplace actions must be included in an IAM policy to allow a role to request or modify access to Bedrock foundation models?"
},
{
"role": "assistant",
"content": "The policy must contain the three actions `aws-marketplace:Subscribe`, `aws-marketplace:U... |
[
{
"role": "user",
"content": "How can I restrict the `aws-marketplace:Subscribe` permission so a user can only request access to a specific model, such as Anthropic Claude 3 Opus?"
},
{
"role": "assistant",
"content": "Add a condition to the `Subscribe` statement using the `aws-marketplace:Produ... |
[
{
"role": "user",
"content": "If I cannot remove request access for Titan, Nova, DeepSeek, Mistral AI, or Meta Llama 3 Instruct, how can I still prevent users from calling these models?"
},
{
"role": "assistant",
"content": "Create an explicit **Deny** statement for inference actions (`bedrock:I... |
[
{
"role": "user",
"content": "Where can I find the product ID for a Bedrock model like Anthropic Claude 3 Opus to use in an IAM policy?"
},
{
"role": "assistant",
"content": "The product IDs are listed in the documentation table. For Anthropic Claude 3 Opus the product ID is `prod-fm3feywmwerog`... |
[
{
"role": "user",
"content": "How do I write an IAM policy that allows a role to subscribe to Anthropic Claude 3 Haiku but denies any inference calls to the Amazon Titan model?"
},
{
"role": "assistant",
"content": "Combine an Allow block for `Subscribe` with the `aws-marketplace:ProductId` cond... |
[
{
"role": "user",
"content": "If I omit the `aws-marketplace:Unsubscribe` action from my IAM policy, will I still be able to revoke a user’s access to a model they previously subscribed to?"
},
{
"role": "assistant",
"content": "No. Without `aws-marketplace:Unsubscribe` the role cannot call the ... |
[
{
"role": "user",
"content": "My application must support German and Japanese in addition to English—should I use the Standard or Classic tier?"
},
{
"role": "assistant",
"content": "Use the Standard tier. It provides \"Extensive language support\" while the Classic tier is limited to \"English,... |
[
{
"role": "user",
"content": "What will happen if my denied‑topic definition exceeds the allowed character limit for a tier?"
},
{
"role": "assistant",
"content": "The guardrail will reject the definition. Classic tier allows a maximum of 200 characters per definition, whereas Standard tier allo... |
[
{
"role": "user",
"content": "Does the Classic tier support cross‑Region inference?"
},
{
"role": "assistant",
"content": "No. The Classic tier does **not** support cross‑Region inference. Only the Standard tier has \"Cross‑Region inference | Supported\"."
}
] |
[
{
"role": "user",
"content": "How can I migrate an existing guardrail from Classic to Standard tier without causing downtime?"
},
{
"role": "assistant",
"content": "1. Modify your guardrail to use the Standard tier and enable cross‑Region inference. 2. (Recommended) Deploy the updated guardrail ... |
[
{
"role": "user",
"content": "Are safeguard tiers available in AWS GovCloud (US‑West)?"
},
{
"role": "assistant",
"content": "Yes. The list of supported regions for safeguard tiers includes \"AWS GovCloud (US-West)\", so both Standard and Classic tiers can be used there."
}
] |
[
{
"role": "user",
"content": "If I change a guardrail’s tier to Standard, do all my other guardrails automatically switch to Standard as well?"
},
{
"role": "assistant",
"content": "No. Changing the tier is per‑guardrail. You must modify each guardrail individually (or automate the update) to us... |
[
{
"role": "user",
"content": "How does the larger 1,000‑character denied‑topic definition limit in the Standard tier improve content filtering compared to the 200‑character limit in Classic?"
},
{
"role": "assistant",
"content": "A longer definition lets you describe more nuanced or multiple pro... |
[
{
"role": "user",
"content": "Why must I enable multi‑factor authentication (MFA) for the AWS root user before using Amazon Bedrock, and how do I do it?"
},
{
"role": "assistant",
"content": "MFA protects the root account—the only identity with unrestricted access to all AWS services. Bedrock re... |
[
{
"role": "user",
"content": "Is the AmazonBedrockFullAccess managed policy sufficient for granting model‑access permissions, or do I also need a custom policy?"
},
{
"role": "assistant",
"content": "The AmazonBedrockFullAccess managed policy grants general Bedrock service permissions, but it do... |
[
{
"role": "user",
"content": "What are the implications of selecting **Enable all models** versus **Enable specific models** when requesting model access?"
},
{
"role": "assistant",
"content": "Enabling **all models** sends a request for every available foundation model, which can lengthen the a... |
[
{
"role": "user",
"content": "After I submit a model‑access request, how can I tell when it has been approved, and how long does it usually take?"
},
{
"role": "assistant",
"content": "Return to the **Model access** page in the Bedrock console. The **Access status** column will change from “Pend... |
[
{
"role": "user",
"content": "How do I switch to the Amazon Bedrock IAM role in the AWS Management Console before requesting model access?"
},
{
"role": "assistant",
"content": "1. Sign in to the AWS Management Console with your IAM Identity Center user. 2. Click your user name at the top right ... |
[
{
"role": "user",
"content": "Is it mandatory to request model access in the US East (N. Virginia) region, and can I later use the models from another AWS region?"
},
{
"role": "assistant",
"content": "Yes. The initial model‑access request must be made while you are in the **US East (N. Virginia... |
[
{
"role": "user",
"content": "Once I have model access, where can I test prompts quickly, and how do I start using the Bedrock API or SDKs?"
},
{
"role": "assistant",
"content": "You can use the **Playgrounds** inside the Bedrock console to run prompts interactively without writing code. For pro... |
[
{
"role": "user",
"content": "Why should I enable fine‑grained access control for my OpenSearch domain if it’s only optional?"
},
{
"role": "assistant",
"content": "Enabling FGAC lets you use role‑based access control, so you can create a narrowly scoped OpenSearch role and map it to the Knowled... |
[
{
"role": "user",
"content": "Which cluster permissions must I add to the OpenSearch role for a Bedrock Knowledge Base to work?"
},
{
"role": "assistant",
"content": "Add the following three cluster permissions:\n- `indices:data/read/msearch`\n- `indices:data/write/bulk*`\n- `indices:data/read/m... |
[
{
"role": "user",
"content": "What index‑level actions should be included in the OpenSearch role to give a Knowledge Base full CRUD access to a vector index?"
},
{
"role": "assistant",
"content": "Add the following index permissions (typically grouped into an action group such as `KnowledgeBases... |
[
{
"role": "user",
"content": "How do I map a custom IAM role versus the automatically created Knowledge Base service role to the OpenSearch role?"
},
{
"role": "assistant",
"content": "1. **Custom IAM role**: Open the IAM console, locate your role, and copy its ARN (e.g., `arn:aws:iam::<accountI... |
[
{
"role": "user",
"content": "What is the correct format of the OpenSearch Dashboards URL to access the Security → Roles page for my domain?"
},
{
"role": "assistant",
"content": "Use the domain endpoint followed by `/ _dashboards/` (note the leading underscore). The full URL looks like:\n`https... |
[
{
"role": "user",
"content": "What will happen if I forget to include `indices:admin/mapping/put` in the OpenSearch role?"
},
{
"role": "assistant",
"content": "The Knowledge Base will be unable to create or update the index mapping for the vector index. Any operation that tries to modify the in... |
[
{
"role": "user",
"content": "Can I reuse the same OpenSearch role for multiple Knowledge Bases, and how should I map each service role?"
},
{
"role": "assistant",
"content": "Yes. A single OpenSearch role can be shared across multiple Knowledge Bases because the role defines the permissions on ... |
[
{
"role": "user",
"content": "What is the visual and functional difference between a data connection and a conditional connection in Bedrock Flows?"
},
{
"role": "assistant",
"content": "A data connection is drawn between the output of one node (the source node) and the input of another node (th... |
[
{
"role": "user",
"content": "How does an expression relate to the ‘whole input’ when defining a node’s individual inputs?"
},
{
"role": "assistant",
"content": "The whole input is the entire payload that arrives from the previous node. An expression defines how to extract a specific part of tha... |
[
{
"role": "user",
"content": "What effect does a branch—created by sending an output to multiple downstream nodes or by using a condition node—have on the final flow invocation response?"
},
{
"role": "assistant",
"content": "When an output is sent to more than one downstream node, or when a con... |
[
{
"role": "user",
"content": "Can a single node have both data connections and conditional connections leaving it, and how are those configured in the Flow builder?"
},
{
"role": "assistant",
"content": "Yes. A node can emit a data connection (solid gray line) to a downstream node for unconditio... |
[
{
"role": "user",
"content": "What is the minimum wiring requirement for a node’s output in a Bedrock Flow?"
},
{
"role": "assistant",
"content": "Each output must be connected to at least one input in a downstream node. The output appears as a circle on the right edge of the node, and you must ... |
[
{
"role": "user",
"content": "How do I define an expression for a node’s input using the Flow builder UI?"
},
{
"role": "assistant",
"content": "In the Flow builder, each input appears as a circle on the left edge of a node. Click that input circle, then select “Add expression”. In the expressio... |
[
{
"role": "user",
"content": "If I connect one node’s output to two downstream nodes, does each downstream node receive a separate copy of the data, and can they modify it independently?"
},
{
"role": "assistant",
"content": "Yes. A data connection sends the same whole input downstream to each t... |
[
{
"role": "user",
"content": "Can I use an existing IAM role for my Bedrock Knowledge Base, and what do I need to change in the OpenSearch domain policy to make it work?"
},
{
"role": "assistant",
"content": "Yes. If you use an existing IAM role, you must ensure the OpenSearch domain’s access po... |
[
{
"role": "user",
"content": "What specific OpenSearch API actions must be allowed in the domain's resource‑based policy for a Bedrock Knowledge Base to function?"
},
{
"role": "assistant",
"content": "The domain policy must allow the following actions: `es:ESHttpPost` (insert information), `es:... |
[
{
"role": "user",
"content": "How do I update the OpenSearch domain’s resource‑based policy to grant the required Bedrock permissions using the AWS Management Console?"
},
{
"role": "assistant",
"content": "1. Open the **Amazon OpenSearch Service console**. 2. Select your domain and go to **Secu... |
[
{
"role": "user",
"content": "If I let Amazon Bedrock create a new service role for the Knowledge Base, what must I do to the OpenSearch domain policy so the role can access the index?"
},
{
"role": "assistant",
"content": "Even when Bedrock auto‑creates the service role, you must manually add t... |
[
{
"role": "user",
"content": "Why do I need to grant `es:ESHttpGet` and `es:ESHttpHead` permissions at both the `domain/index` and `domain/index/*` levels?"
},
{
"role": "assistant",
"content": "Granting at `domain/index` allows Bedrock to retrieve high‑level metadata about the index (such as en... |
[
{
"role": "user",
"content": "What happens if my OpenSearch domain has a restrictive access policy, and how can I resolve it so Bedrock Knowledge Bases can access the index?"
},
{
"role": "assistant",
"content": "A restrictive policy can prevent the Bedrock service role (whether existing or auto... |
[
{
"role": "user",
"content": "When I choose “Create and use a new service role” in the Bedrock console, can I specify a custom role name, and does that affect how I write the OpenSearch resource policy?"
},
{
"role": "assistant",
"content": "Yes. After selecting **Create and use a new service ro... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.