Add support for local llm (mistral 7b) (#31)

This commit is contained in:
Adil Hafeez 2024-08-06 23:40:06 -07:00 committed by GitHub
parent b49fc2f264
commit 445b1ea210
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
24 changed files with 703 additions and 51 deletions

View file

@ -0,0 +1,22 @@
# Weather forecasting
This demo shows how you can use intelligent prompt gateway to provide realtime weather forecast using Mistral LLM locally hosted using llama.cpp as LLM Hosting Service.
# Startig the demo
1. Ensure that submodule is up to date
```sh
git submodule sync --recursive
```
1. Download mistral 7b model using following shell command
```sh
sh download_mistral_7b.sh
```
2. Start services
```sh
docker compose up
```
3. Navigate to http://localhost:18080/
4. You can type in queries like "how is the weather in Seattle"
1. You can also ask follow up questions like "show me sunny days"
5. To see metrics navigate to "http://localhost:3000/" (use admin/grafana for login)
1. Open up dahsboard named "Intelligent Gateway Overview"
2. On this dashboard you can see reuqest latency and number of requests

View file

@ -0,0 +1,88 @@
services:
config-generator:
build:
context: ../../
dockerfile: config_generator/Dockerfile
volumes:
- ./katanemo-config.yaml:/usr/src/app/katanemo-config.yaml
- ./generated:/usr/src/app/out
envoy:
build:
context: ../../
dockerfile: envoyfilter/Dockerfile
hostname: envoy
ports:
- "10000:10000"
- "19901:9901"
volumes:
- ./generated/envoy.yaml:/etc/envoy/envoy.yaml
- /etc/ssl/cert.pem:/etc/ssl/cert.pem
depends_on:
config-generator:
condition: service_completed_successfully
embeddingserver:
condition: service_healthy
embeddingserver:
build:
context: ../../embedding-server
dockerfile: Dockerfile
ports:
- "18081:80"
healthcheck:
test: ["CMD", "curl" ,"http://localhost:80/healthz"]
interval: 5s
retries: 20
volumes:
- ~/.cache/huggingface:/root/.cache/huggingface
qdrant:
image: qdrant/qdrant
hostname: vector-db
ports:
- 16333:6333
- 16334:6334
chatbot-ui:
build:
context: ../../chatbot-ui
dockerfile: Dockerfile
ports:
- "18080:8080"
environment:
- OPENAI_API_KEY=${OPENAI_API_KEY}
- CHAT_COMPLETION_ENDPOINT=http://envoy:10000/v1/chat/completions
prometheus:
image: prom/prometheus
container_name: prometheus
command:
- '--config.file=/etc/prometheus/prometheus.yaml'
ports:
- 9090:9090
restart: unless-stopped
volumes:
- ./prometheus:/etc/prometheus
- ./prom_data:/prometheus
grafana:
image: grafana/grafana
container_name: grafana
ports:
- 3000:3000
restart: unless-stopped
environment:
- GF_SECURITY_ADMIN_USER=admin
- GF_SECURITY_ADMIN_PASSWORD=grafana
volumes:
- ./grafana:/etc/grafana/provisioning/datasources
- ./grafana/dashboard.yaml:/etc/grafana/provisioning/dashboards/main.yaml
- ./grafana/dashboards:/var/lib/grafana/dashboards
mistral_7b_instruct:
image: ghcr.io/ggerganov/llama.cpp:server
hostname: mistral_7b_instruct
ports:
- "10001:10001"
volumes:
- ./mistral-7b-instruct-v0.2.Q4_K_M.gguf:/models/model.gguf
command: ["--host", "0.0.0.0", "--port", "10001", "-m", "/models/model.gguf"]

View file

@ -0,0 +1 @@
huggingface-cli download TheBloke/Mistral-7B-Instruct-v0.2-GGUF mistral-7b-instruct-v0.2.Q4_K_M.gguf --local-dir . --local-dir-use-symlinks False

View file

@ -0,0 +1,12 @@
apiVersion: 1
providers:
- name: "Dashboard provider"
orgId: 1
type: file
disableDeletion: false
updateIntervalSeconds: 10
allowUiUpdates: false
options:
path: /var/lib/grafana/dashboards
foldersFromFilesStructure: true

View file

@ -0,0 +1,355 @@
{
"annotations": {
"list": [
{
"builtIn": 1,
"datasource": {
"type": "grafana",
"uid": "-- Grafana --"
},
"enable": true,
"hide": true,
"iconColor": "rgba(0, 211, 255, 1)",
"name": "Annotations & Alerts",
"type": "dashboard"
}
]
},
"editable": true,
"fiscalYearStartMonth": 0,
"graphTooltip": 1,
"links": [],
"panels": [
{
"datasource": {
"type": "prometheus",
"uid": "PBFA97CFB590B2093"
},
"fieldConfig": {
"defaults": {
"color": {
"mode": "palette-classic"
},
"custom": {
"axisBorderShow": false,
"axisCenteredZero": false,
"axisColorMode": "text",
"axisLabel": "",
"axisPlacement": "auto",
"barAlignment": 0,
"drawStyle": "line",
"fillOpacity": 0,
"gradientMode": "none",
"hideFrom": {
"legend": false,
"tooltip": false,
"viz": false
},
"insertNulls": false,
"lineInterpolation": "linear",
"lineWidth": 1,
"pointSize": 5,
"scaleDistribution": {
"type": "linear"
},
"showPoints": "auto",
"spanNulls": false,
"stacking": {
"group": "A",
"mode": "none"
},
"thresholdsStyle": {
"mode": "off"
}
},
"mappings": [],
"thresholds": {
"mode": "absolute",
"steps": [
{
"color": "green",
"value": null
},
{
"color": "red",
"value": 80
}
]
}
},
"overrides": []
},
"gridPos": {
"h": 8,
"w": 12,
"x": 0,
"y": 0
},
"id": 2,
"options": {
"legend": {
"calcs": [],
"displayMode": "list",
"placement": "bottom",
"showLegend": true
},
"tooltip": {
"mode": "single",
"sort": "none"
}
},
"targets": [
{
"datasource": {
"type": "prometheus",
"uid": "PBFA97CFB590B2093"
},
"disableTextWrap": false,
"editorMode": "code",
"expr": "avg(rate(envoy_cluster_internal_upstream_rq_time_sum[1m]) / rate(envoy_cluster_internal_upstream_rq_time_count[1m])) by (envoy_cluster_name)",
"fullMetaSearch": false,
"hide": false,
"includeNullMetadata": true,
"instant": false,
"legendFormat": "__auto",
"range": true,
"refId": "A",
"useBackend": false
}
],
"title": "request latency - internal (ms)",
"type": "timeseries"
},
{
"datasource": {
"type": "prometheus",
"uid": "PBFA97CFB590B2093"
},
"fieldConfig": {
"defaults": {
"color": {
"mode": "palette-classic"
},
"custom": {
"axisBorderShow": false,
"axisCenteredZero": false,
"axisColorMode": "text",
"axisLabel": "",
"axisPlacement": "auto",
"barAlignment": 0,
"drawStyle": "line",
"fillOpacity": 0,
"gradientMode": "none",
"hideFrom": {
"legend": false,
"tooltip": false,
"viz": false
},
"insertNulls": false,
"lineInterpolation": "linear",
"lineWidth": 1,
"pointSize": 5,
"scaleDistribution": {
"type": "linear"
},
"showPoints": "auto",
"spanNulls": false,
"stacking": {
"group": "A",
"mode": "none"
},
"thresholdsStyle": {
"mode": "off"
}
},
"mappings": [],
"thresholds": {
"mode": "absolute",
"steps": [
{
"color": "green",
"value": null
},
{
"color": "red",
"value": 80
}
]
}
},
"overrides": []
},
"gridPos": {
"h": 8,
"w": 12,
"x": 12,
"y": 0
},
"id": 1,
"options": {
"legend": {
"calcs": [],
"displayMode": "list",
"placement": "bottom",
"showLegend": true
},
"tooltip": {
"mode": "single",
"sort": "none"
}
},
"targets": [
{
"datasource": {
"type": "prometheus",
"uid": "PBFA97CFB590B2093"
},
"disableTextWrap": false,
"editorMode": "code",
"expr": "avg(rate(envoy_cluster_external_upstream_rq_time_sum[1m]) / rate(envoy_cluster_external_upstream_rq_time_count[1m])) by (envoy_cluster_name)",
"fullMetaSearch": false,
"hide": false,
"includeNullMetadata": true,
"instant": false,
"legendFormat": "__auto",
"range": true,
"refId": "A",
"useBackend": false
}
],
"title": "request latency - external (ms)",
"type": "timeseries"
},
{
"datasource": {
"type": "prometheus",
"uid": "PBFA97CFB590B2093"
},
"fieldConfig": {
"defaults": {
"color": {
"mode": "palette-classic"
},
"custom": {
"axisBorderShow": false,
"axisCenteredZero": false,
"axisColorMode": "text",
"axisLabel": "",
"axisPlacement": "auto",
"barAlignment": 0,
"drawStyle": "line",
"fillOpacity": 0,
"gradientMode": "none",
"hideFrom": {
"legend": false,
"tooltip": false,
"viz": false
},
"insertNulls": false,
"lineInterpolation": "linear",
"lineWidth": 1,
"pointSize": 5,
"scaleDistribution": {
"type": "linear"
},
"showPoints": "auto",
"spanNulls": false,
"stacking": {
"group": "A",
"mode": "none"
},
"thresholdsStyle": {
"mode": "off"
}
},
"mappings": [],
"thresholds": {
"mode": "absolute",
"steps": [
{
"color": "green",
"value": null
},
{
"color": "red",
"value": 80
}
]
}
},
"overrides": []
},
"gridPos": {
"h": 8,
"w": 12,
"x": 0,
"y": 8
},
"id": 3,
"options": {
"legend": {
"calcs": [],
"displayMode": "list",
"placement": "bottom",
"showLegend": true
},
"tooltip": {
"mode": "single",
"sort": "none"
}
},
"targets": [
{
"datasource": {
"type": "prometheus",
"uid": "PBFA97CFB590B2093"
},
"disableTextWrap": false,
"editorMode": "code",
"expr": "avg(rate(envoy_cluster_internal_upstream_rq_completed[1m])) by (envoy_cluster_name)",
"fullMetaSearch": false,
"includeNullMetadata": true,
"instant": false,
"legendFormat": "__auto",
"range": true,
"refId": "A",
"useBackend": false
},
{
"datasource": {
"type": "prometheus",
"uid": "PBFA97CFB590B2093"
},
"disableTextWrap": false,
"editorMode": "code",
"expr": "avg(rate(envoy_cluster_external_upstream_rq_completed[1m])) by (envoy_cluster_name)",
"fullMetaSearch": false,
"hide": false,
"includeNullMetadata": true,
"instant": false,
"legendFormat": "__auto",
"range": true,
"refId": "B",
"useBackend": false
}
],
"title": "Upstream request count",
"type": "timeseries"
}
],
"schemaVersion": 39,
"tags": [],
"templating": {
"list": []
},
"time": {
"from": "now-15m",
"to": "now"
},
"timepicker": {},
"timezone": "browser",
"title": "Intelligent Gateway Overview",
"uid": "adt6uhx5lk8aob",
"version": 3,
"weekStart": ""
}

View file

@ -0,0 +1,9 @@
apiVersion: 1
datasources:
- name: Prometheus
type: prometheus
url: http://prometheus:9090
isDefault: true
access: proxy
editable: true

View file

@ -0,0 +1,46 @@
default_prompt_endpoint: "127.0.0.1"
load_balancing: "round_robin"
timeout_ms: 5000
embedding_provider:
name: "SentenceTransformer"
model: "all-MiniLM-L6-v2"
llm_providers:
- name: open-ai-gpt-4
api_key: "$OPEN_AI_API_KEY"
model: gpt-4
- name: mistral_7b_instruct
model: mistral-7b-instruct
endpoint: http://mistral_7b_instruct:10001/v1/chat/completions
default: true
prompt_targets:
- type: context_resolver
name: weather_forecast
few_shot_examples:
- what is the weather in New York?
- how is the weather in San Francisco?
- what is the forecast in Chicago?
entities:
- name: city
required: true
- name: days
endpoint:
cluster: weatherhost
path: /weather
system_prompt: |
You are a helpful weather forecaster. Use weater data that is provided to you. Please following following guidelines when responding to user queries:
- Use farenheight for temperature
- Use miles per hour for wind speed
#TODO: add support for adding custom clusters e.g.
# clusters:
# qdrant:
# options:
# - address: "qdrant"
# - address: "weatherhost"
# - port: 6333

View file

@ -0,0 +1,23 @@
global:
scrape_interval: 15s
scrape_timeout: 10s
evaluation_interval: 15s
alerting:
alertmanagers:
- static_configs:
- targets: []
scheme: http
timeout: 10s
api_version: v1
scrape_configs:
- job_name: envoy
honor_timestamps: true
scrape_interval: 15s
scrape_timeout: 10s
metrics_path: /stats
scheme: http
static_configs:
- targets:
- envoy:9901
params:
format: ['prometheus']