[BUG] Failed to get data source ID: No data sources found for knowledge base INSBJQX888

[trey-rosius]

Checks

• I have updated to the lastest minor and patch version of Strands
• I have checked the documentation and this is not expected behavior
• I have searched ./issues and there are no duplicates of my issue

Strands Version

0.1.6

Python Version

3.13

Operating System

macOS 15.3.2

Installation Method

pip

Steps to Reproduce

1. Install Strands
2. Run the application

Expected Behavior

I expect the agent to retrieve information from the bedrock knowledge base

Actual Behavior

response is {'status': 'error', 'content': [{'text': '❌ Failed to get data source ID: No data sources found for knowledge base INSBJQX4KN'}], 'toolUseId': 'tooluse_memory_402927007'}

Additional Context

I have the knowledge base already created in Bedrock

Here's the complete code

import os

from strands import Agent
from strands_tools import http_request
from strands_tools import use_llm, memory
from strands.models import  BedrockModel

STRANDS_KNOWLEDGE_BASE_ID=os.environ["STRANDS_KNOWLEDGE_BASE_ID"]
SYSTEM_PROMPT = """
You are an AI Social Media Post Generation Agent. Your job is to take a user’s request and create a ready-to-post message for the specified social media platform. You have access to a knowledge base for additional context; if the user’s topic requires background or factual information, retrieve relevant details from the knowledge base before composing the post.

──────────────
### Capabilities
1. **Multi-Platform**  
   • The user will specify a target platform (e.g., Twitter, Facebook, LinkedIn, Instagram).  
   • Adhere to each platform’s best practices and tone (e.g., LinkedIn is more professional, Instagram can be more visual and hashtag‐heavy).

2. **Twitter Character Limit Enforcement**  
   • If the user requests a Twitter post, ensure the final message does not exceed 280 characters (including spaces and punctuation).  
   • Use URL shorteners or trim unnecessary words if needed to stay within limits.  
   • If given a long piece of content, summarize concisely without losing the core message.

3. **Knowledge Base Querying**  
   • Before drafting, consult the knowledge base for relevant facts, statistics, quotes, or background information about the topic.  
   • If the user explicitly references the knowledge base (“fetch statistics on X”), gather those details first and weave them into the post.  
   • If no knowledge base content is needed, you may skip this step and rely on general knowledge.

4. **Tone and Style Adaptation**  
   • Match the user’s requested tone (e.g., friendly, professional, humorous, motivational).  
   • Use appropriate hashtags, emojis, or formatting according to platform norms.  
   • Keep posts engaging, concise, and clear.

──────────────
### Input Schema
Users will send a JSON-like request with fields:
• `platform`: required.  
• `topic`: required.  
• `tone`: if omitted, default to “engaging” and “informal.”  
• `length`: if “short,” prioritize brevity; if “long,” you may use up to the platform’s maximum.  
• `useKB`: if true, first query knowledge base for relevant context before writing.

──────────────
### Output Requirements
• Return exactly one text string: the post ready for publishing.  
• Do not wrap the text in quotes or metadata.  
• Do not include analysis, reasoning, or commentary—only the final post content.  
• If the platform is X(Twitter), verify the character count ≤ 280. If you exceed 280 characters, revise until it fits.

──────────────
### Processing Steps
1. **Validate `platform`**:  
   - If `platform` is “X(Twitter),” prepare to count characters.  
   - If unknown, ask the user to specify one of the supported platforms.

2. **Optionally Query Knowledge Base**:  
   - If `"useKB": true` or the topic seems to require extra facts, search the knowledge base for up-to-date information.  
   - Use only the most relevant 1–2 sentences from the KB to inform your post.

3. **Compose the Post**:  
   - Apply platform‐specific best practices (e.g., hashtags for Instagram, no more than 2–3 hashtags on X(Twitter), professional tone on LinkedIn).  
   - Use the user’s requested tone and length.  
   - For X(Twitter), ensure ≤ 280 characters. Count “https://…” or “@username” properly as part of the character limit.  
   - For other platforms, respect reasonable length guidelines (e.g., 150–200 words maximum for Facebook or LinkedIn).

4. **Final Check**:  
   - If X(Twitter): recount characters, revise if needed.  
   - Ensure the final post reads naturally, is engaging, and accurately reflects the topic.  
   - Do not mention KB details explicitly; weave them seamlessly into the narrative.


"""
bedrock_model = BedrockModel(
    model_id="us.anthropic.claude-3-7-sonnet-20250219-v1:0",
    region_name='us-east-1',
    temperature=0.3,
    cache_tools='default',

)
# Create an agent with HTTP capabilities
weather_agent = Agent(
    model=bedrock_model,
    system_prompt=SYSTEM_PROMPT,
    tools=[use_llm,memory],  # Explicitly enable http_request tool
)
result = weather_agent.tool.memory(action="retrieve",
                    STRANDS_KNOWLEDGE_BASE_ID="INSBJQX4KN",
                    query="Create a post on the course 'Build A Nanny Booking API' for linkedIn", min_score=0.4, max_results=9)
print(f"response is {result}")

Possible Solution

No response

Related Issues

No response

[mrityunjayshukla]

Hi @trey-rosius , did u add any of data sources to your knowledge base? Ref: https://docs.aws.amazon.com/bedrock/latest/userguide/data-source-connectors.html

[trey-rosius]

Hi @trey-rosius , did u add any of data sources to your knowledge base? Ref: https://docs.aws.amazon.com/bedrock/latest/userguide/data-source-connectors.html

yes i did. Actually, i tried with 2 different knowledge bases. One that's currently in use by another application. It has an S3 bucket as the datasource and a pinecone vector store.
I also created a new Knowledgebase (S3 datasource + pinecone). I added files to the datasource. Synced. Tested the Knowledge base on the AWS Console and it works properly.
Then tried again through Strands and same error

response is {'status': 'error', 'content': [{'text': '❌ Failed to get data source ID: No data sources found for knowledge base 0VAFFPBUTY'}], 'toolUseId': 'tooluse_memory_481452797'}

[mrityunjayshukla]

can you share output for AWS CLI command aws bedrock-agent list-knowledge-base-data-sources \ --knowledge-base-id INSBJQX4KN or for KB Id = 0VAFFPBUTY using same AWS credential that agent is using? Just want to confirm that direct API call is able to return the data sources associated with Bedrock KB?

[trey-rosius]

can you share output for AWS CLI command aws bedrock-agent list-knowledge-base-data-sources \ --knowledge-base-id INSBJQX4KN or for KB Id = 0VAFFPBUTY using same AWS credential that agent is using? Just want to confirm that direct API call is able to return the data sources associated with Bedrock KB?

For this command

 aws bedrock-agent list-data-sources --knowledge-base-id 0VAFFPBUTY

Here's the response

{
    "dataSourceSummaries": [
        {
            "dataSourceId": "MZFORIAYMI",
            "knowledgeBaseId": "0VAFFPBUTY",
            "name": "knowledge-base-quick-start-az318-data-source",
            "status": "AVAILABLE",
            "updatedAt": "2025-06-08T14:21:10.792933+00:00"
        }
    ]
}

[mrityunjayshukla]

I see. can you explicitly set AWS_REGION = us-east-1 as env variable and retry once?

[trey-rosius]

I see. can you explicitly set AWS_REGION = us-east-1 as env variable and retry once?

yeppppp... that did the trick. Thanks. Should i close this ?

[mrityunjayshukla]

yes, @trey-rosius . If you don't configure aws region via environment variable then you need to specify provide region in tool e.g.

# Call memory tool with explicit region parameter
kb_response = agent.tool.memory(
    action="retrieve", 
    query="Create a post on the course 'Build A Nanny Booking API' for linkedIn"
    max_results=3,
    region="us-east-1"  # Specify the region directly in the tool call
)

Otherwise default aws region in strands ai-agent is us-west-2 region.

@awsarron / @zastrowm , I think we should include aws region name in error messages. Also, tagging you in case we need to update our documentation regarding the same.

[zastrowm]

@mrityunjayshukla agreed that region would be helpful in error messages - also curious if we should throw if no region is specified (or set via env variable) to make that more obvious - rather than defaulting to us-west-2 which seems arbitrary.

I'd be curious what the team thinks about special casing the bedrock-client region for aws-tools.

Also I'll consider this as part of strands-agents/tools#31 - there's a question of how default values (like regions) should be set and if they're changeable, which might change how some of this is figured out.