Skip to content

Commit

Permalink
Merge branch 'dev' into docs
Browse files Browse the repository at this point in the history
  • Loading branch information
@ksopyla
ksopyla committed Jun 16, 2024
2 parents 7db5a40 + ddd0ae0 commit eedca0b
Show file tree
Hide file tree
Showing 100 changed files with 9,444 additions and 2,871 deletions.
92 changes: 86 additions & 6 deletions backend/provider_price_list.json
View file
Original file line number Diff line number Diff line change
Expand Up @@ -220,14 +220,74 @@
"is_active": true
},
{
"model_name": "gemini-pro",
"provider": "Google Vertex AI",
"model_name": "gemini-1.0-pro-001",
"provider": "Google VertexAI",
"start_date": "",
"match_pattern": "(?i)^(gemini-pro)(@[a-zA-Z0-9]+)?$",
"input_price": 0.00025,
"output_price": 0.0005,
"match_pattern": "(?i)^(gemini-1.0-pro-001)?$",
"input_price": 0.000125,
"output_price": 0.000375,
"total_price": 0,
"is_active": false
"is_active": true
},
{
"model_name": "gemini-1.0-pro-002",
"provider": "Google VertexAI",
"start_date": "",
"match_pattern": "(?i)^(gemini-1.0-pro-002)?$",
"input_price": 0.000125,
"output_price": 0.000375,
"total_price": 0,
"is_active": true
},
{
"model_name": "gemini-1.0-pro-vision-001",
"provider": "Google VertexAI",
"start_date": "",
"match_pattern": "(?i)^(gemini-1.0-pro-vision-001)?$",
"input_price": 0.000125,
"output_price": 0.000375,
"total_price": 0,
"is_active": true
},
{
"model_name": "gemini-1.5-flash-001",
"provider": "Google VertexAI",
"start_date": "",
"match_pattern": "(?i)^(gemini-1.5-flash-001)?$",
"input_price": 0.000125,
"output_price": 0.000375,
"total_price": 0,
"is_active": true
},
{
"model_name": "gemini-1.5-flash-preview-0514",
"provider": "Google VertexAI",
"start_date": "",
"match_pattern": "(?i)^(gemini-1.5-flash-preview-0514)?$",
"input_price": 0.000125,
"output_price": 0.000375,
"total_price": 0,
"is_active": true
},
{
"model_name": "gemini-1.5-pro-001",
"provider": "Google VertexAI",
"start_date": "",
"match_pattern": "(?i)^(gemini-1.5-pro-001)?$",
"input_price": 0.00125,
"output_price": 0.00375,
"total_price": 0,
"is_active": true
},
{
"model_name": "gemini-1.5-pro-preview-0514",
"provider": "Google VertexAI",
"start_date": "",
"match_pattern": "(?i)^(gemini-1.5-pro-preview-0514)?$",
"input_price": 0.00125,
"output_price": 0.00375,
"total_price": 0,
"is_active": true
},
{
"model_name": "gpt-35-turbo-0125",
Expand Down Expand Up @@ -479,6 +539,26 @@
"total_price": 0,
"is_active": true
},
{
"model_name": "gpt-4o-2024-05-13",
"provider": "Azure OpenAI",
"start_date": "",
"match_pattern": "(?i)^(gpt-4o)(|-2024-05-13)$",
"input_price": 0.005,
"output_price": 0.015,
"total_price": 0,
"is_active": true
},
{
"model_name": "gpt-4o-2024-05-13",
"provider": "OpenAI",
"start_date": "",
"match_pattern": "(?i)^(gpt-4o)(|-2024-05-13)$",
"input_price": 0.005,
"output_price": 0.015,
"total_price": 0,
"is_active": true
},
{
"model_name": "text-bison",
"provider": "Google Vertex AI",
Expand Down
12 changes: 0 additions & 12 deletions backend/pyproject.toml
View file
Original file line number Diff line number Diff line change
Expand Up @@ -35,18 +35,6 @@ requests-mock = "^1.11.0"
locust = "^2.24.1"


[tool.poetry.group.examples]
optional = true

[tool.poetry.group.examples.dependencies]
ipykernel = "^6.25.2"
langchain = "^0.0"
openai = "^1"
boto3 = "^1.34.13"
anthropic = "^0.25.6"
rich = "^13.7.1"
langchain-anthropic = "^0.1.11"

[build-system]
requires = ["poetry-core>=1.0.0"]
build-backend = "poetry.core.masonry.api"
117 changes: 111 additions & 6 deletions backend/src/app/web_api.py
View file
Original file line number Diff line number Diff line change
@@ -1,10 +1,13 @@
import re
from typing import Annotated, Any

import utils
from _datetime import datetime, timezone
from app.dependencies import get_provider_pricelist, get_transaction_context
from auth.authorization import decode_and_validate_token
from auth.schemas import GetUserSchema
from auth.models import User
from auth.schemas import GetPartialUserSchema, GetUserSchema
from auth.use_cases import get_all_users
from fastapi import Depends, Request, Security
from fastapi.responses import JSONResponse
from lato import TransactionContext
Expand All @@ -27,8 +30,10 @@
from slugify import slugify
from transactions.models import generate_uuid
from transactions.schemas import (
CreateTransactionSchema,
GetTransactionLatencyStatisticsWithoutDateSchema,
GetTransactionPageResponseSchema,
GetTransactionSchema,
GetTransactionsLatencyStatisticsSchema,
GetTransactionStatusStatisticsSchema,
GetTransactionsUsageStatisticsSchema,
Expand All @@ -37,6 +42,7 @@
StatisticTransactionSchema,
)
from transactions.use_cases import (
add_transaction,
count_token_usage_for_project,
count_transactions,
delete_multiple_transactions,
Expand All @@ -51,7 +57,7 @@

@app.get("/api/auth/whoami", dependencies=[Security(decode_and_validate_token)])
def whoami(
request: Request, user: dict = Depends(decode_and_validate_token)
request: Request, user: User = Depends(decode_and_validate_token)
) -> GetUserSchema:
return GetUserSchema(
external_id=user.external_id,
Expand Down Expand Up @@ -256,6 +262,9 @@ async def get_paginated_transactions(
project_id: str | None = None,
sort_field: str | None = None,
sort_type: str | None = None,
status_codes: str | None = None,
providers: str | None = None,
models: str | None = None,
) -> GetTransactionPageResponseSchema:
"""
API endpoint to retrieve a paginated list of transactions based on specified filters.
Expand All @@ -269,9 +278,18 @@ async def get_paginated_transactions(
:param project_id: Optional. Project ID to filter transactions by.
:param sort_field: Optional. Field to sort by.
:param sort_type: Optional. Ordering method (asc or desc).
:param status_codes: Optional. List of status codes for filtering transactions.
:param providers: Optional. List of providers for filtering transactions.
:param models: Optional. List of models for filtering transactions.
"""
if tags is not None:
tags = tags.split(",")
if status_codes is not None:
status_codes = list(map(lambda x: int(x), status_codes.split(",")))
if providers is not None:
providers = providers.split(",")
if models is not None:
models = models.split(",")

transactions = ctx.call(
get_all_filtered_and_paginated_transactions,
Expand All @@ -283,6 +301,9 @@ async def get_paginated_transactions(
project_id=project_id,
sort_field=sort_field,
sort_type=sort_type,
status_codes=status_codes,
providers=providers,
models=models,
)

projects = ctx.call(get_all_projects)
Expand All @@ -305,6 +326,9 @@ async def get_paginated_transactions(
date_from=date_from,
date_to=date_to,
project_id=project_id,
status_codes=status_codes,
providers=providers,
models=models,
)
page_response = GetTransactionPageResponseSchema(
items=new_transactions,
Expand Down Expand Up @@ -355,7 +379,7 @@ async def get_transaction_usage_statistics_over_time(
project_id=project_id,
date_from=date_from,
date_to=date_to,
status_code=200,
status_codes=[200],
)
if count == 0:
return []
Expand All @@ -365,7 +389,7 @@ async def get_transaction_usage_statistics_over_time(
project_id=project_id,
date_from=date_from,
date_to=date_to,
status_code=200,
status_codes=[200],
)
transactions = [
StatisticTransactionSchema(
Expand Down Expand Up @@ -527,7 +551,7 @@ async def get_transaction_latency_statistics_over_time(
project_id=project_id,
date_from=date_from,
date_to=date_to,
status_code=200,
status_codes=[200],
null_generation_speed=False,
)
if count == 0:
Expand All @@ -538,7 +562,7 @@ async def get_transaction_latency_statistics_over_time(
project_id=project_id,
date_from=date_from,
date_to=date_to,
status_code=200,
status_codes=[200],
null_generation_speed=False,
)
transactions = [
Expand Down Expand Up @@ -640,6 +664,87 @@ async def get_config(
}


@app.get("/api/users", dependencies=[Security(decode_and_validate_token)])
async def get_users(
ctx: Annotated[TransactionContext, Depends(get_transaction_context)],
auth_user: User = Depends(decode_and_validate_token),
) -> list[GetPartialUserSchema]:
users = ctx.call(get_all_users)
idx = next(
(i for i, usr in enumerate(users) if usr.external_id == auth_user.external_id),
None,
)
if idx is not None:
temp = users.pop(idx)
users.insert(0, temp)
else:
users.insert(0, User(**auth_user.model_dump()))

parsed_users = map(
lambda user: GetPartialUserSchema(
id=user.id,
email=user.email,
full_name=str(user.given_name + " " + user.family_name),
picture=user.picture,
),
users,
)
return list(parsed_users)


@app.post(
"/api/transactions",
response_class=JSONResponse,
status_code=201,
dependencies=[Security(decode_and_validate_token)],
)
def create_transaction(
request: Request,
data: CreateTransactionSchema,
ctx: Annotated[TransactionContext, Depends(get_transaction_context)],
) -> GetTransactionSchema:
if ((data.status_code == 200) and data.model and data.provider) and not (
data.input_cost or data.output_cost or data.total_cost
):
pricelist = get_provider_pricelist(request)
pricelist = [
item
for item in pricelist
if item.provider == data.provider
and re.match(item.match_pattern, data.model)
]
if len(pricelist) > 0:
if pricelist[0].input_price == 0:
data.input_cost, data.output_cost = 0, 0
data.total_cost = (
(data.input_tokens + data.output_tokens)
/ 1000
* pricelist[0].total_price
)
else:
data.input_cost = pricelist[0].input_price * (data.input_tokens / 1000)
data.output_cost = pricelist[0].output_price * (
data.output_tokens / 1000
)
data.total_cost = data.input_cost + data.output_cost
else:
data.input_cost, data.output_cost, data.total_cost = None, None, None

if not data.generation_speed:
if data.output_tokens is not None and (data.output_tokens > 0):
data.generation_speed = (
data.output_tokens
/ (datetime.now(tz=timezone.utc) - data.request_time).total_seconds()
)
elif data.output_tokens == 0:
data.generation_speed = None
else:
data.generation_speed = 0

created_transaction = ctx.call(add_transaction, data=data)
return GetTransactionSchema(**created_transaction.model_dump())


@app.post("/api/only_for_purpose/mock_transactions", response_class=JSONResponse)
async def mock_transactions(
ctx: Annotated[TransactionContext, Depends(get_transaction_context)],
Expand Down
Loading

0 comments on commit eedca0b

Please sign in to comment.