mirror of
https://github.com/FlipsideCrypto/livequery-models.git
synced 2026-02-06 10:56:46 +00:00
remove groq
This commit is contained in:
Jensen Yap
parent
1e0e73069b
commit
53772e8d3d
@ -1,265 +0,0 @@
|
||||
# Groq API Integration
|
||||
|
||||
This directory contains Snowflake UDFs for integrating with the Groq API, providing fast inference with various open-source language models.
|
||||
|
||||
## Available Models
|
||||
|
||||
- **llama3-8b-8192**: Meta Llama 3 8B model with 8K context (Very Fast)
|
||||
- **llama3-70b-8192**: Meta Llama 3 70B model with 8K context (Fast, better quality)
|
||||
- **gemma-7b-it**: Google Gemma 7B instruction-tuned (Instruction following)
|
||||
|
||||
**Note**: Check [Groq's documentation](https://console.groq.com/docs/models) for the latest available models, or query the live model list with:
|
||||
|
||||
```sql
|
||||
-- Get current list of available models
|
||||
SELECT groq_utils.list_models();
|
||||
|
||||
-- Get details about a specific model
|
||||
SELECT groq_utils.get_model_info('llama3-8b-8192');
|
||||
```
|
||||
|
||||
## Setup
|
||||
|
||||
1. Get your Groq API key from [https://console.groq.com/keys](https://console.groq.com/keys)
|
||||
|
||||
2. Store the API key in Snowflake secrets:
|
||||
- **System users**: Store under `_FSC_SYS/GROQ`
|
||||
- **Regular users**: Store under `vault/groq/api`
|
||||
|
||||
3. Deploy the Groq marketplace functions:
|
||||
```bash
|
||||
dbt run --models groq__ groq_utils__groq_utils
|
||||
```
|
||||
|
||||
**Note**: Groq functions automatically use the appropriate secret path based on your user type.
|
||||
|
||||
## Functions
|
||||
|
||||
### `groq.chat_completions(messages, [model], [max_tokens], [temperature], [top_p], [frequency_penalty], [presence_penalty])`
|
||||
|
||||
Send messages to Groq for chat completion.
|
||||
|
||||
### `groq.quick_chat(user_message, [system_message])`
|
||||
|
||||
Quick single or system+user message chat.
|
||||
|
||||
**Note**: All functions use the `GROQ_API_KEY` environment variable for authentication.
|
||||
|
||||
### `groq.extract_response_text(groq_response)`
|
||||
|
||||
Extract text content from Groq API responses.
|
||||
|
||||
### `groq_utils.post_api(path, body)`
|
||||
|
||||
Low-level HTTP POST to Groq API endpoints.
|
||||
|
||||
### `groq_utils.get_api(path)`
|
||||
|
||||
Low-level HTTP GET to Groq API endpoints.
|
||||
|
||||
### `groq_utils.list_models()`
|
||||
|
||||
List all available models from Groq API.
|
||||
|
||||
### `groq_utils.get_model_info(model_id)`
|
||||
|
||||
Get information about a specific model.
|
||||
|
||||
## Examples
|
||||
|
||||
### Basic Chat
|
||||
```sql
|
||||
-- Simple chat with default model (llama3-8b-8192)
|
||||
SELECT groq.chat_completions(
|
||||
[{'role': 'user', 'content': 'Explain quantum computing in simple terms'}]
|
||||
);
|
||||
|
||||
-- Quick chat shorthand
|
||||
SELECT groq.quick_chat('What is the capital of France?');
|
||||
```
|
||||
|
||||
### Chat with System Prompt
|
||||
```sql
|
||||
-- Chat with system prompt using quick_chat
|
||||
SELECT groq.quick_chat(
|
||||
'You are a helpful Python programming assistant.',
|
||||
'How do I create a list comprehension?'
|
||||
);
|
||||
|
||||
-- Full chat_completions with system message
|
||||
SELECT groq.chat_completions(
|
||||
[
|
||||
{'role': 'system', 'content': 'You are a data scientist expert.'},
|
||||
{'role': 'user', 'content': 'Explain the difference between supervised and unsupervised learning'}
|
||||
]
|
||||
);
|
||||
```
|
||||
|
||||
### Different Models
|
||||
```sql
|
||||
-- Use the larger, more capable model
|
||||
SELECT groq.chat_completions(
|
||||
[{'role': 'user', 'content': 'Write a Python function to calculate fibonacci numbers'}],
|
||||
'llama3-70b-8192',
|
||||
500 -- max_tokens
|
||||
);
|
||||
|
||||
-- Use the larger model for better quality
|
||||
SELECT groq.chat_completions(
|
||||
[{'role': 'user', 'content': 'Analyze this complex problem...'}],
|
||||
'llama3-70b-8192'
|
||||
);
|
||||
```
|
||||
|
||||
### Custom Parameters
|
||||
```sql
|
||||
-- Fine-tune response generation
|
||||
SELECT groq.chat_completions(
|
||||
[{'role': 'user', 'content': 'Generate creative story ideas'}],
|
||||
'llama3-8b-8192', -- model
|
||||
300, -- max_tokens
|
||||
0.8, -- temperature (more creative)
|
||||
0.9, -- top_p
|
||||
0.1, -- frequency_penalty (reduce repetition)
|
||||
0.1 -- presence_penalty (encourage new topics)
|
||||
);
|
||||
```
|
||||
|
||||
|
||||
### Extract Response Text
|
||||
```sql
|
||||
-- Get just the text content from API response
|
||||
WITH chat_response AS (
|
||||
SELECT groq.quick_chat('Hello, how are you?') as response
|
||||
)
|
||||
SELECT groq.extract_response_text(response) as message_text
|
||||
FROM chat_response;
|
||||
```
|
||||
|
||||
### Conversational Chat
|
||||
```sql
|
||||
-- Multi-turn conversation
|
||||
SELECT groq.chat_completions([
|
||||
{'role': 'system', 'content': 'You are a helpful coding assistant.'},
|
||||
{'role': 'user', 'content': 'I need help with SQL queries'},
|
||||
{'role': 'assistant', 'content': 'I\'d be happy to help with SQL! What specific query are you working on?'},
|
||||
{'role': 'user', 'content': 'How do I join two tables with a LEFT JOIN?'}
|
||||
]);
|
||||
```
|
||||
|
||||
### Model Comparison
|
||||
```sql
|
||||
-- Compare responses from different models
|
||||
WITH responses AS (
|
||||
SELECT
|
||||
'llama3-8b-8192' as model,
|
||||
groq.extract_response_text(
|
||||
groq.chat_completions([{'role': 'user', 'content': 'Explain machine learning'}], 'llama3-8b-8192', 100)
|
||||
) as response
|
||||
UNION ALL
|
||||
SELECT
|
||||
'llama3-70b-8192' as model,
|
||||
groq.extract_response_text(
|
||||
groq.chat_completions([{'role': 'user', 'content': 'Explain machine learning'}], 'llama3-70b-8192', 100)
|
||||
) as response
|
||||
)
|
||||
SELECT * FROM responses;
|
||||
```
|
||||
|
||||
### Batch Processing
|
||||
```sql
|
||||
-- Process multiple questions
|
||||
WITH questions AS (
|
||||
SELECT * FROM VALUES
|
||||
('What is Python?'),
|
||||
('What is JavaScript?'),
|
||||
('What is SQL?')
|
||||
AS t(question)
|
||||
)
|
||||
SELECT
|
||||
question,
|
||||
groq.extract_response_text(
|
||||
groq.quick_chat(question, 'You are a programming tutor.')
|
||||
) as answer
|
||||
FROM questions;
|
||||
```
|
||||
|
||||
### Get Available Models
|
||||
```sql
|
||||
-- List all available models with details
|
||||
SELECT
|
||||
model.value:id::STRING as model_id,
|
||||
model.value:object::STRING as object_type,
|
||||
model.value:created::INTEGER as created_timestamp,
|
||||
model.value:owned_by::STRING as owned_by
|
||||
FROM (
|
||||
SELECT groq_utils.list_models() as response
|
||||
),
|
||||
LATERAL FLATTEN(input => response:data) as model
|
||||
ORDER BY model_id;
|
||||
|
||||
-- Check if a specific model is available
|
||||
WITH models AS (
|
||||
SELECT groq_utils.list_models() as response
|
||||
)
|
||||
SELECT
|
||||
CASE
|
||||
WHEN ARRAY_CONTAINS('llama3-70b-8192'::VARIANT, response:data[*]:id)
|
||||
THEN 'Model is available'
|
||||
ELSE 'Model not found'
|
||||
END as availability
|
||||
FROM models;
|
||||
```
|
||||
|
||||
### GitHub Actions Integration Example
|
||||
```sql
|
||||
-- Example of how this is used in GitHub Actions failure analysis
|
||||
SELECT
|
||||
run_id,
|
||||
groq.extract_response_text(
|
||||
groq.quick_chat(
|
||||
CONCAT('Analyze this failure: Job=', job_name, ' Error=', error_logs),
|
||||
'You are analyzing CI/CD failures. Provide concise root cause analysis.'
|
||||
)
|
||||
) as ai_analysis
|
||||
FROM my_failed_jobs
|
||||
WHERE run_id = '12345678';
|
||||
```
|
||||
|
||||
## Error Handling
|
||||
|
||||
The functions include built-in error handling. Check for errors in responses:
|
||||
|
||||
```sql
|
||||
WITH response AS (
|
||||
SELECT groq.quick_chat('Hello') as result
|
||||
)
|
||||
SELECT
|
||||
CASE
|
||||
WHEN result:error IS NOT NULL THEN result:error:message::STRING
|
||||
ELSE groq.extract_response_text(result)
|
||||
END as final_response
|
||||
FROM response;
|
||||
```
|
||||
|
||||
## Performance Tips
|
||||
|
||||
1. **Model Selection**: Use `llama3-8b-8192` for fast, simple tasks and `llama3-70b-8192` for complex reasoning
|
||||
2. **Token Limits**: Set appropriate `max_tokens` to control costs and response length
|
||||
3. **Temperature**: Use lower values (0.1-0.3) for factual tasks, higher (0.7-1.0) for creative tasks
|
||||
4. **Stay Updated**: Check Groq's model documentation regularly as they add new models and deprecate others
|
||||
|
||||
## Integration with GitHub Actions
|
||||
|
||||
This Groq integration is used by the GitHub Actions failure analysis system in `slack_notify` macro:
|
||||
|
||||
```sql
|
||||
-- In your GitHub Actions workflow
|
||||
dbt run-operation slack_notify --vars '{
|
||||
"owner": "your-org",
|
||||
"repo": "your-repo",
|
||||
"run_id": "12345678",
|
||||
"ai_provider": "groq",
|
||||
"enable_ai_analysis": true
|
||||
}'
|
||||
```
|
||||
@ -1,55 +0,0 @@
|
||||
{% macro config_groq_chat_udfs(schema_name = "groq", utils_schema_name = "groq_utils") -%}
|
||||
{#
|
||||
This macro is used to generate API calls to Groq chat completion endpoints
|
||||
#}
|
||||
|
||||
- name: {{ schema_name -}}.chat_completions
|
||||
signature:
|
||||
- [MESSAGES, ARRAY, Array of message objects]
|
||||
- [MODEL, STRING, The model to use (optional, defaults to llama3-8b-8192)]
|
||||
return_type:
|
||||
- "VARIANT"
|
||||
options: |
|
||||
COMMENT = $$Send messages to Groq and get a chat completion response with optional model selection [API docs: Chat Completions](https://console.groq.com/docs/api-reference#chat-completions)$$
|
||||
sql: |
|
||||
SELECT groq_utils.post_api(
|
||||
'/openai/v1/chat/completions',
|
||||
{
|
||||
'model': COALESCE(MODEL, 'llama3-8b-8192'),
|
||||
'messages': MESSAGES,
|
||||
'max_tokens': 1024,
|
||||
'temperature': 0.1
|
||||
}
|
||||
) as response
|
||||
|
||||
- name: {{ schema_name -}}.quick_chat
|
||||
signature:
|
||||
- [USER_MESSAGE, STRING, The user message to send]
|
||||
- [MODEL, STRING, The model to use (optional, defaults to llama3-8b-8192)]
|
||||
return_type:
|
||||
- "VARIANT"
|
||||
options: |
|
||||
COMMENT = $$Quick single message chat with Groq using optional model selection$$
|
||||
sql: |
|
||||
SELECT {{ schema_name }}.chat_completions(
|
||||
ARRAY_CONSTRUCT(
|
||||
OBJECT_CONSTRUCT('role', 'user', 'content', USER_MESSAGE)
|
||||
),
|
||||
MODEL
|
||||
) as response
|
||||
|
||||
- name: {{ schema_name -}}.extract_response_text
|
||||
signature:
|
||||
- [GROQ_RESPONSE, VARIANT, The response object from Groq API]
|
||||
return_type:
|
||||
- "STRING"
|
||||
options: |
|
||||
COMMENT = $$Extract the text content from a Groq chat completion response$$
|
||||
sql: |
|
||||
SELECT COALESCE(
|
||||
GROQ_RESPONSE:choices[0]:message:content::STRING,
|
||||
GROQ_RESPONSE:error:message::STRING,
|
||||
'No response available'
|
||||
)
|
||||
|
||||
{% endmacro %}
|
||||
@ -1,66 +0,0 @@
|
||||
{% macro config_groq_utils_udfs(schema_name = "groq_utils", utils_schema_name = "groq_utils") -%}
|
||||
{#
|
||||
This macro is used to generate API calls to Groq API endpoints
|
||||
#}
|
||||
- name: {{ schema_name -}}.post_api
|
||||
signature:
|
||||
- [PATH, STRING, The API endpoint path]
|
||||
- [BODY, OBJECT, The request body]
|
||||
return_type:
|
||||
- "VARIANT"
|
||||
options: |
|
||||
COMMENT = $$Make POST requests to Groq API [API docs: Groq](https://console.groq.com/docs/api-reference)$$
|
||||
sql: |
|
||||
SELECT live.udf_api_v2(
|
||||
'POST',
|
||||
CONCAT('https://api.groq.com', PATH),
|
||||
{
|
||||
'Authorization': 'Bearer {API_KEY}',
|
||||
'Content-Type': 'application/json',
|
||||
'fsc-quantum-execution-mode': 'async'
|
||||
},
|
||||
BODY,
|
||||
IFF(_utils.udf_whoami() <> CURRENT_USER(), '_FSC_SYS/GROQ', 'Vault/prod/livequery/groq')
|
||||
) as response
|
||||
|
||||
- name: {{ schema_name -}}.get_api
|
||||
signature:
|
||||
- [PATH, STRING, The API endpoint path]
|
||||
return_type:
|
||||
- "VARIANT"
|
||||
options: |
|
||||
COMMENT = $$Make GET requests to Groq API [API docs: Groq](https://console.groq.com/docs/api-reference)$$
|
||||
sql: |
|
||||
SELECT live.udf_api_v2(
|
||||
'GET',
|
||||
CONCAT('https://api.groq.com', PATH),
|
||||
{
|
||||
'Authorization': 'Bearer {API_KEY}',
|
||||
'Content-Type': 'application/json',
|
||||
'fsc-quantum-execution-mode': 'async'
|
||||
},
|
||||
NULL,
|
||||
IFF(_utils.udf_whoami() <> CURRENT_USER(), '_FSC_SYS/GROQ', 'Vault/prod/livequery/groq')
|
||||
) as response
|
||||
|
||||
- name: {{ schema_name -}}.list_models
|
||||
signature:
|
||||
- []
|
||||
return_type:
|
||||
- "VARIANT"
|
||||
options: |
|
||||
COMMENT = $$List available models from Groq API$$
|
||||
sql: |
|
||||
SELECT {{ schema_name }}.get('/openai/v1/models')
|
||||
|
||||
- name: {{ schema_name -}}.get_model_info
|
||||
signature:
|
||||
- [MODEL_ID, STRING, The model ID to get info for]
|
||||
return_type:
|
||||
- "VARIANT"
|
||||
options: |
|
||||
COMMENT = $$Get information about a specific model$$
|
||||
sql: |
|
||||
SELECT {{ schema_name }}.get_api('/openai/v1/models/' || MODEL_ID)
|
||||
|
||||
{% endmacro %}
|
||||
@ -1,6 +0,0 @@
|
||||
-- depends_on: {{ ref('live') }}
|
||||
-- depends_on: {{ ref('groq_utils__groq_utils') }}
|
||||
{%- set configs = [
|
||||
config_groq_chat_udfs,
|
||||
] -%}
|
||||
{{- ephemeral_deploy_marketplace(configs) -}}
|
||||
@ -1,41 +0,0 @@
|
||||
version: 2
|
||||
models:
|
||||
- name: groq__
|
||||
columns:
|
||||
- name: chat_completions
|
||||
tests:
|
||||
- test_udf_without_context:
|
||||
name: test_groq__chat_completions_simple
|
||||
args: >
|
||||
[{'role': 'user', 'content': 'Hello, how are you?'}],
|
||||
'llama3-8b-8192'
|
||||
assertions:
|
||||
- result:data:choices IS NOT NULL
|
||||
- result:data:model IS NOT NULL
|
||||
|
||||
- name: quick_chat
|
||||
tests:
|
||||
- test_udf_without_context:
|
||||
name: test_groq__quick_chat_single_message
|
||||
args: >
|
||||
'Hello, how are you?',
|
||||
'llama3-8b-8192'
|
||||
assertions:
|
||||
- result:data:choices IS NOT NULL
|
||||
|
||||
|
||||
- name: extract_response_text
|
||||
tests:
|
||||
- test_udf_without_context:
|
||||
name: test_groq__extract_response_text
|
||||
args: >
|
||||
{'choices': [{'message': {'content': 'Hello there!'}}]}
|
||||
assertions:
|
||||
- result = 'Hello there!'
|
||||
- test_udf_without_context:
|
||||
name: test_groq__extract_response_text_error
|
||||
args: >
|
||||
{'error': {'message': 'API Error occurred'}}
|
||||
assertions:
|
||||
- result = 'API Error occurred'
|
||||
|
||||
@ -1,5 +0,0 @@
|
||||
-- depends_on: {{ ref('live') }}
|
||||
{%- set configs = [
|
||||
config_groq_utils_udfs,
|
||||
] -%}
|
||||
{{- ephemeral_deploy_marketplace(configs) -}}
|
||||
Loading…
Reference in New Issue
Block a user