Configure Amazon Bedrock

Amazon Bedrock is a fully managed foundation model service provided by AWS, integrating top-tier models from Anthropic Claude, Meta Llama, Amazon Titan, and more. It offers enterprise-grade security, privacy protection, and flexible model selection, supporting pay-as-you-go pricing.

1. Obtain AWS Bedrock Access Credentials

1.1 Visit AWS Console

Visit AWS Management Console and log in: https://console.aws.amazon.com/

1.2 Enter Bedrock Service

1. Enter Bedrock in the search box
2. Click Amazon Bedrock
3. Select your region (recommended us-east-1)

1.3 Request Model Access

1. Click Model access in the left menu
2. Click Manage model access
3. Check the models you want to use (e.g., Claude 3.5 Sonnet, Llama 3.1, Mistral Large)
4. Click Request model access
5. Wait for approval (usually within a few minutes)

1.4 Create IAM User and Access Keys

1. Visit IAM console: https://console.aws.amazon.com/iam/
2. Click Users > Add user
3. Enter username (e.g., cuemate-bedrock)
4. Select Access key - Programmatic access
5. Attach policy: AmazonBedrockFullAccess
6. Complete creation and record Access Key ID and Secret Access Key

2. Configure Bedrock Model in CueMate

2.1 Enter Model Settings Page

After logging into CueMate, click Model Settings in the dropdown menu in the upper right corner.

2.2 Add New Model

Click the Add Model button in the upper right corner.

2.3 Select Amazon Bedrock Provider

In the popup dialog:

1. Provider Type: Select Amazon Bedrock
2. After clicking, it will automatically proceed to the next step

2.4 Fill in Configuration Information

Fill in the following information on the configuration page:

Basic Configuration

1. Model Name: Give this model configuration a name (e.g., Bedrock Claude 4.5 Sonnet)
2. API Key: Paste the long-term API key obtained from AWS Bedrock
3. AWS Region: Select region (e.g., us-east-1, default value)
4. Model Version: Select or enter the model ID you want to use. Common models include:
   • anthropic.claude-sonnet-4-5-20250929-v1:0: Claude 4.5 Sonnet (latest, max output 64K)
   • anthropic.claude-haiku-4-5-20251001-v1:0: Claude 4.5 Haiku (fast, max output 64K)
   • anthropic.claude-opus-4-1-20250805-v1:0: Claude 4.1 Opus (high performance, max output 32K)
   • anthropic.claude-3-5-sonnet-20241022-v2:0: Claude 3.5 Sonnet (max output 64K)
   • amazon.nova-pro-v1:0: Amazon Nova Pro (multimodal, max output 8K)
   • meta.llama3-1-405b-instruct-v1:0: Llama 3.1 405B (ultra-large scale, max output 8K)
   • mistral.mistral-large-2407-v1:0: Mistral Large (high performance, max output 8K)

Advanced Configuration (Optional)

Expand the Advanced Configuration panel to adjust the following parameters:

CueMate Interface Adjustable Parameters:

1. Temperature: Controls output randomness

   • Range: 0-2 (depending on model series)
   • Recommended Value: 0.7
   • Function: Higher values produce more random and creative output, lower values produce more stable and conservative output
   • Model Ranges:
     • Claude series: 0-1
     • Llama series: 0-2
     • Mistral series: 0-1
     • DeepSeek series: 0-2
     • Amazon Titan series: 0-1
   • Usage Suggestions:
     • Creative writing/brainstorming: 0.8-1.2 (based on model upper limit)
     • Regular conversation/Q&A: 0.6-0.8
     • Code generation/precise tasks: 0.3-0.5

2. Max Tokens: Limits single output length

   • Range: 256 - 65536 (depending on model)
   • Recommended Value: 8192
   • Function: Controls the maximum word count of model's single response
   • Model Limits:
     • Claude 4.5 Sonnet/Haiku: max 65536 (64K tokens)
     • Claude 4 Opus: max 32768 (32K tokens)
     • Claude 4 Sonnet: max 65536 (64K tokens)
     • Claude 3.7 Sonnet: max 65536 (64K tokens)
     • Claude 3.5 Sonnet: max 65536 (64K tokens)
     • Claude 3.5 Haiku: max 8192 (8K tokens)
     • Claude 3 Opus/Sonnet/Haiku: max 4096 (4K tokens)
     • Amazon Nova all series: max 8192 (8K tokens)
     • Amazon Titan all series: max 8192 (8K tokens)
     • Meta Llama all series: max 8192 (8K tokens)
     • Mistral all series: max 8192 (8K tokens)
     • DeepSeek all series: max 8192 (8K tokens)
     • AI21 Jamba series: max 4096 (4K tokens)
     • Cohere Command series: max 4096 (4K tokens)
     • Qwen all series: max 8192 (8K tokens)
   • Usage Suggestions:
     • Short Q&A: 1024-2048
     • Regular conversation: 4096-8192
     • Long text generation: 16384-32768
     • Ultra-long documents: 65536 (Claude 4.5/4/3.7/3.5 Sonnet only)

Other Advanced Parameters Supported by AWS Bedrock API:

Although CueMate's interface only provides temperature and max_tokens adjustments, if you call AWS Bedrock directly through the API, different model series also support the following advanced parameters:

Anthropic Claude Series Parameters

3. top_p (nucleus sampling)

   • Range: 0-1
   • Default Value: 1
   • Function: Samples from the smallest candidate set with cumulative probability reaching p
   • Relationship with temperature: Usually only adjust one of them
   • Usage Suggestions:
     • Maintain diversity: 0.9-0.95
     • More conservative output: 0.7-0.8

4. top_k

   • Range: 0-500
   • Default Value: 250
   • Function: Samples from the k candidates with highest probability
   • Usage Suggestions:
     • More diverse: 300-500
     • More conservative: 50-150

5. stop_sequences (stop sequence)

   • Type: String array
   • Default Value: ["\n\nHuman:"]
   • Function: Stops when generated content contains specified string
   • Maximum Quantity: 4
   • Example: ["###", "User:", "\n\n"]

Meta Llama Series Parameters

6. top_p (nucleus sampling)

   • Range: 0-1
   • Default Value: 0.9
   • Function: Samples from the smallest candidate set with cumulative probability reaching p

7. top_k

   • Range: 1-500
   • Default Value: 50
   • Function: Samples from the k candidates with highest probability

Amazon Titan Series Parameters

8. topP (nucleus sampling)

   • Range: 0-1
   • Default Value: 1
   • Function: Samples from the smallest candidate set with cumulative probability reaching p

9. stopSequences (stop sequence)

   • Type: String array
   • Function: Stops when generated content contains specified string
   • Example: ["User:", "###"]

Mistral Series Parameters

10. top_p (nucleus sampling)

    • Range: 0-1
    • Default Value: 1
    • Function: Samples from the smallest candidate set with cumulative probability reaching p

11. top_k

    • Range: 0-200
    • Default Value: 50
    • Function: Samples from the k candidates with highest probability

AWS Bedrock General Features:

12. stream

    • Type: Boolean
    • Default Value: false
    • Function: Enables streaming return, returning as it generates
    • In CueMate: Handled automatically, no manual setting required

13. guardrails (safety guardrails)

    • Type: Object
    • Function: Configure AWS Bedrock Guardrails for content filtering
    • Usage Scenario: Enterprise-level security compliance requirements
    • Example:

      {
        "guardrailIdentifier": "your-guardrail-id",
        "guardrailVersion": "1"
      }

Scenario	Model Series	temperature	max_tokens	top_p	top_k
Creative writing	Claude	0.8-0.9	4096-8192	0.95	300
Code generation	Claude	0.3-0.5	2048-4096	0.9	100
Q&A system	Claude	0.7	1024-2048	0.9	250
Complex reasoning	Claude Opus	0.7	32768	0.9	250
Long text generation	Claude Sonnet	0.7	65536	0.9	250
Fast response	Claude Haiku	0.6	4096	0.9	200
Large-scale reasoning	Llama 3.1 405B	0.7	8192	0.9	50
Multimodal tasks	Nova Pro	0.7	8192	1.0	-

2.5 Test Connection

After filling in the configuration, click the Test Connection button to verify if the configuration is correct.

If the configuration is correct, it will display a success message and return a model response example.

If the configuration is incorrect, it will display test error logs, and you can view specific error information through log management.

2.6 Save Configuration

After a successful test, click the Save button to complete the model configuration.

3. Use Model

Through the dropdown menu in the upper right corner, enter the system settings interface and select the model configuration you want to use in the Large Model Provider section.

After configuration, you can select to use this model in functions such as interview training and question generation, or you can individually select the model configuration for a specific interview in the interview options.

4. Supported Model List

CueMate supports all mainstream large models on the AWS Bedrock platform, covering 70+ models from 10+ providers. Below are representative models from each series:

4.1 Anthropic Claude Series

No.	Model Name	Model ID	Max Output	Use Case
1	Claude 4.5 Sonnet	anthropic.claude-sonnet-4-5-20250929-v1:0	64K tokens	Latest generation, general scenarios, high-performance reasoning
2	Claude 4.5 Haiku	anthropic.claude-haiku-4-5-20251001-v1:0	64K tokens	Fast response, high throughput
3	Claude 4.1 Opus	anthropic.claude-opus-4-1-20250805-v1:0	32K tokens	Complex reasoning, deep analysis
4	Claude 4 Opus	anthropic.claude-opus-4-20250514-v1:0	32K tokens	High-quality output
5	Claude 4 Sonnet	anthropic.claude-sonnet-4-20250514-v1:0	64K tokens	Balanced performance and cost
6	Claude 3.7 Sonnet	anthropic.claude-3-7-sonnet-20250219-v1:0	64K tokens	Enhanced general model
7	Claude 3.5 Sonnet v2	anthropic.claude-3-5-sonnet-20241022-v2:0	64K tokens	General scenarios, high performance
8	Claude 3.5 Sonnet v1	anthropic.claude-3-5-sonnet-20240620-v1:0	64K tokens	General scenarios
9	Claude 3.5 Haiku	anthropic.claude-3-5-haiku-20241022-v1:0	8K tokens	Fast response
10	Claude 3 Opus	anthropic.claude-3-opus-20240229-v1:0	4K tokens	Complex reasoning
11	Claude 3 Sonnet	anthropic.claude-3-sonnet-20240229-v1:0	4K tokens	Balanced performance
12	Claude 3 Haiku	anthropic.claude-3-haiku-20240307-v1:0	4K tokens	Lightweight tasks

4.2 Amazon Nova Series

No.	Model Name	Model ID	Max Output	Use Case
1	Nova Premier	amazon.nova-premier-v1:0	8K tokens	Flagship multimodal model
2	Nova Pro	amazon.nova-pro-v1:0	8K tokens	High-performance multimodal processing
3	Nova Lite	amazon.nova-lite-v1:0	8K tokens	Lightweight multimodal tasks
4	Nova Micro	amazon.nova-micro-v1:0	8K tokens	Ultra-lightweight scenarios
5	Nova Sonic	amazon.nova-sonic-v1:0	8K tokens	Fast response

4.3 Amazon Titan Series

No.	Model Name	Model ID	Max Output	Use Case
1	Titan Premier	amazon.titan-text-premier-v1:0	8K tokens	Enterprise applications
2	Titan Express	amazon.titan-text-express-v1	8K tokens	Fast response
3	Titan Lite	amazon.titan-text-lite-v1	8K tokens	Lightweight tasks

4.4 Meta Llama Series

No.	Model Name	Model ID	Max Output	Use Case
1	Llama 4 Scout 17B	meta.llama4-scout-17b-instruct-v1:0	8K tokens	New generation medium-scale model
2	Llama 4 Maverick 17B	meta.llama4-maverick-17b-instruct-v1:0	8K tokens	New generation high-performance model
3	Llama 3.3 70B	meta.llama3-3-70b-instruct-v1:0	8K tokens	Enhanced large-scale reasoning
4	Llama 3.2 90B	meta.llama3-2-90b-instruct-v1:0	8K tokens	Large-scale reasoning
5	Llama 3.2 11B	meta.llama3-2-11b-instruct-v1:0	8K tokens	Medium-scale tasks
6	Llama 3.2 3B	meta.llama3-2-3b-instruct-v1:0	8K tokens	Lightweight tasks
7	Llama 3.2 1B	meta.llama3-2-1b-instruct-v1:0	8K tokens	Ultra-lightweight
8	Llama 3.1 405B	meta.llama3-1-405b-instruct-v1:0	8K tokens	Ultra-large scale reasoning
9	Llama 3.1 70B	meta.llama3-1-70b-instruct-v1:0	8K tokens	Large-scale tasks
10	Llama 3.1 8B	meta.llama3-1-8b-instruct-v1:0	8K tokens	Standard tasks
11	Llama 3 70B	meta.llama3-70b-instruct-v1:0	8K tokens	Classic large-scale model
12	Llama 3 8B	meta.llama3-8b-instruct-v1:0	8K tokens	Classic standard model

4.5 Mistral AI Series

No.	Model Name	Model ID	Max Output	Use Case
1	Pixtral Large 2502	mistral.pixtral-large-2502-v1:0	8K tokens	Multimodal large model
2	Mistral Large 2407	mistral.mistral-large-2407-v1:0	8K tokens	High-performance scenarios
3	Mistral Large 2402	mistral.mistral-large-2402-v1:0	8K tokens	General scenarios
4	Mistral Small 2402	mistral.mistral-small-2402-v1:0	8K tokens	Lightweight and efficient
5	Mixtral 8x7B	mistral.mixtral-8x7b-instruct-v0:1	4K tokens	Mixture of experts model
6	Mistral 7B	mistral.mistral-7b-instruct-v0:2	8K tokens	Lightweight tasks

4.6 AI21 Labs Series

No.	Model Name	Model ID	Max Output	Use Case
1	Jamba 1.5 Large	ai21.jamba-1-5-large-v1:0	4K tokens	Large-scale hybrid architecture
2	Jamba 1.5 Mini	ai21.jamba-1-5-mini-v1:0	4K tokens	Lightweight hybrid architecture

4.7 Cohere Series

No.	Model Name	Model ID	Max Output	Use Case
1	Command R+	cohere.command-r-plus-v1:0	4K tokens	Enhanced command model
2	Command R	cohere.command-r-v1:0	4K tokens	Standard command model

4.8 DeepSeek Series

No.	Model Name	Model ID	Max Output	Use Case
1	DeepSeek R1	deepseek.r1-v1:0	8K tokens	Reasoning optimized version
2	DeepSeek V3	deepseek.v3-v1:0	8K tokens	Latest generation model

4.9 Qwen Series

No.	Model Name	Model ID	Max Output	Use Case
1	Qwen 3 Coder 480B	qwen.qwen3-coder-480b-a35b-v1:0	8K tokens	Ultra-large scale code generation
2	Qwen 3 235B	qwen.qwen3-235b-a22b-2507-v1:0	8K tokens	Ultra-large scale general reasoning
3	Qwen 3 Coder 30B	qwen.qwen3-coder-30b-a3b-v1:0	8K tokens	Code generation specialized
4	Qwen 3 32B	qwen.qwen3-32b-v1:0	8K tokens	Standard general model

4.10 OpenAI Series

No.	Model Name	Model ID	Max Output	Use Case
1	GPT-OSS 120B	openai.gpt-oss-120b-1:0	4K tokens	Open-source GPT large model
2	GPT-OSS 20B	openai.gpt-oss-20b-1:0	4K tokens	Open-source GPT medium model

Notes:

• All models above require requesting access permissions in the AWS Bedrock console before use
• Different models have different pricing, please refer to AWS Bedrock Pricing
• Actual maximum output depends on the Max Tokens parameter you set in CueMate configuration
• Some models may only be available in specific AWS regions, it's recommended to use us-east-1 for best model coverage

5. Common Issues

5.1 Model Access Not Authorized

Symptom: Model access denied prompt

Solution:

1. Check model access status in Bedrock console
2. Confirm that model access has been requested and obtained
3. Wait for approval (some models may require 1-2 business days)

5.2 Insufficient IAM Permissions

Symptom: Permission error prompt

Solution:

1. Confirm that IAM user has attached AmazonBedrockFullAccess policy
2. Check if access keys are correct
3. Verify that region settings match the model's available region

5.3 Region Not Supported

Symptom: Service not available in current region prompt

Solution:

1. Use regions that support Bedrock (recommended us-east-1 or us-west-2)
2. Modify region code in API URL
3. Confirm that the selected model is available in that region

5.4 Quota Limit

Symptom: Exceeds request quota prompt

Solution:

1. Check quota usage in Bedrock console
2. Request to increase TPM (tokens per minute) or RPM (requests per minute) limits
3. Optimize request frequency

Related Links

• AWS Bedrock Console
• Bedrock Documentation
• Pricing Information
• Supported Model List