Merge branch 'dev' into f/sdk-verb-update

.env.example

@@ -8,8 +8,7 @@ COZO_ROCKSDB_DIR=cozo.db
 DTYPE=bfloat16
 EMBEDDING_SERVICE_URL=http://text-embeddings-inference/embed
 GATEWAY_PORT=80
-GENERATION_AUTH_TOKEN=myauthkey
-GENERATION_URL=http://model-serving:8000/v1
+OPENAI_API_KEY=""
 GPU_MEMORY_UTILIZATION=0.95
 HF_TOKEN=""
 HUGGING_FACE_HUB_TOKEN=""
@@ -21,17 +20,17 @@ GF_SECURITY_ADMIN_PASSWORD=changethis
 MODEL_API_KEY=myauthkey
 MODEL_API_KEY_HEADER_NAME=Authorization
 MODEL_API_URL=http://model-serving:8000
+MODEL_INFERENCE_URL=http://model-serving:8000/v1
 MODEL_ID=BAAI/llm-embedder
-MODEL_NAME=julep-ai/samantha-1-turbo
-# MODEL_NAME = "julep-ai/samantha-1-turbo-awq"
+MODEL_NAME = "julep-ai/samantha-1-turbo"
 SKIP_CHECK_DEVELOPER_HEADERS=true
 SUMMARIZATION_TOKENS_THRESHOLD=2048
 TEMPERATURE_SCALING_FACTOR=0.9
 TEMPERATURE_SCALING_POWER=0.9
 TEMPORAL_ENDPOINT=temporal:7233
 TEMPORAL_NAMESPACE=default
 TEMPORAL_WORKER_URL=temporal:7233
-TP_SIZE=2
+TP_SIZE=1
 TRUNCATE_EMBED_TEXT=true
 TRAEFIK_LOG_LEVEL=DEBUG
-WORKER_URL=temporal:7233
+WORKER_URL=temporal:7233

.github/workflows/lint-and-format.yml

@@ -1,19 +1,107 @@
 name: Lint and format APIs and SDKs
 run-name: ${{ github.actor }} is linting and formatting the code
+
 on: [push, pull_request]
+
 jobs:
-  Lint-And-Format-APIs-And-SDKs:
+  Lint-And-Format:
     runs-on: ubuntu-latest
+
+    strategy:
+      matrix:
+        directory: [agents-api, model-serving, sdks/python]
+
     steps:
       - uses: actions/checkout@v4
+
       - uses: actions/setup-python@v5
         with:
-          python-version: '3.10' 
+          python-version: "3.10"
+
       - name: Install and configure Poetry
         uses: snok/install-poetry@v1
-      - name: Lint and format Agents API
-        run: cd agents-api && poetry install && poetry run poe check
-      - name: Lint and format Models API
-        run: cd model-serving && poetry install && poetry run poe check
-      - name: Lint and format Python SDK
-        run: cd sdks/python && poetry install && poetry run poe check
+
+      - name: Configure Poetry to use .venv
+        run: |
+          cd ${{ matrix.directory }}
+          poetry config virtualenvs.in-project true
+
+      - name: Cache Poetry virtualenv
+        uses: actions/cache@v3
+        with:
+          path: ${{ matrix.directory }}/.venv
+          key: ${{ runner.os }}-poetry-${{ matrix.directory }}-${{ hashFiles(format('{0}/poetry.lock', matrix.directory)) }}
+          restore-keys: |
+            ${{ runner.os }}-poetry-${{ matrix.directory }}-
+
+      - name: Cache pytype
+        uses: actions/cache@v3
+        with:
+          path: ${{ matrix.directory }}/.pytype
+          key: ${{ runner.os }}-pytype-${{ matrix.directory }}-${{ hashFiles(format('{0}/**/*.py', matrix.directory)) }}
+          restore-keys: |
+            ${{ runner.os }}-pytype-${{ matrix.directory }}-
+
+      - name: Install dependencies
+        run: |
+          cd ${{ matrix.directory }}
+          poetry install
+
+      - name: Lint and format
+        run: |
+          cd ${{ matrix.directory }}
+          poetry run poe format
+          poetry run poe lint
+
+      - name: Typecheck
+        run: |
+          cd ${{ matrix.directory }}
+          poetry run poe typecheck
+
+      - uses: stefanzweifel/git-auto-commit-action@v4
+        with:
+          commit_message: "refactor: Lint ${{ matrix.directory }} (CI)"
+          branch: ${{ github.head_ref }}
+
+  Lint-And-Format-TS-SDK:
+    runs-on: ubuntu-latest
+
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Use Node.js
+        uses: actions/setup-node@v3
+        with:
+          node-version: "20.x"
+
+      - name: Cache npm dependencies
+        uses: actions/cache@v3
+        with:
+          path: sdks/ts/node_modules
+          key: ${{ runner.os }}-npm-${{ hashFiles('sdks/ts/package-lock.json') }}
+          restore-keys: |
+            ${{ runner.os }}-npm-
+
+      - name: Install npm dependencies
+        run: |
+          cd sdks/ts
+          npm ci
+
+      - name: Lint and format TypeScript SDK
+        run: |
+          cd sdks/ts
+          npm run format
+
+      - name: Build TypeScript SDK
+        run: |
+          cd sdks/ts
+          # npm run build
+
+      - uses: stefanzweifel/git-auto-commit-action@v4
+        with:
+          commit_message: "refactor: Lint sdks/ts (CI)"
+          branch: ${{ github.head_ref }}
+
+concurrency:
+  group: ${{ github.workflow }}-${{ github.ref }}
+  cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}

.gitignore

@@ -6,3 +6,4 @@ ngrok*
 *.pyc
 */node_modules/
 .aider*
+.vscode/

agents-api/agents_api/activities/__init__.py

@@ -11,4 +11,4 @@
 - `summarization.py`: Summarizes dialogues and updates memory based on the conversation context.
 
 This module plays a crucial role in enhancing the capabilities of agents by providing them with the tools to understand and process information more effectively.
-"""
+"""

agents-api/agents_api/activities/co_density.py

@@ -3,7 +3,7 @@
 
 from temporalio import activity
 
-from ..clients.openai import client as openai_client
+from ..clients.model import julep_client
 from .types import MemoryDensityTaskArgs
 
 
@@ -63,7 +63,7 @@ async def run_prompt(
 ) -> str:
     prompt = make_prompt(MemoryDensityTaskArgs(memory=memory))
 
-    response = await openai_client.chat.completions.create(
+    response = await julep_client.chat.completions.create(
         model=model,
         messages=[
             {

agents-api/agents_api/activities/dialog_insights.py

@@ -3,7 +3,7 @@
 
 from temporalio import activity
 
-from ..clients.openai import client as openai_client
+from ..clients.model import julep_client
 from .types import ChatML, DialogInsightsTaskArgs
 
 
@@ -66,7 +66,7 @@ async def run_prompt(
         DialogInsightsTaskArgs(dialog=dialog, person1=person1, person2=person2)
     )
 
-    response = await openai_client.chat.completions.create(
+    response = await julep_client.chat.completions.create(
         model=model,
         messages=[
             {

agents-api/agents_api/activities/mem_mgmt.py

@@ -4,7 +4,7 @@
 
 from temporalio import activity
 
-from ..clients.openai import client as openai_client
+from ..clients.model import julep_client
 from .types import ChatML, MemoryManagementTaskArgs
 
 
@@ -135,7 +135,7 @@ async def run_prompt(
         )
     )
 
-    response = await openai_client.chat.completions.create(
+    response = await julep_client.chat.completions.create(
         model=model,
         messages=[
             {

agents-api/agents_api/activities/mem_rating.py

@@ -3,7 +3,7 @@
 
 from temporalio import activity
 
-from ..clients.openai import client as openai_client
+from ..clients.model import julep_client
 from .types import MemoryRatingTaskArgs
 
 
@@ -47,7 +47,7 @@ async def run_prompt(
 ) -> str:
     prompt = make_prompt(MemoryRatingTaskArgs(memory=memory))
 
-    response = await openai_client.chat.completions.create(
+    response = await julep_client.chat.completions.create(
         model=model,
         messages=[
             {

agents-api/agents_api/activities/relationship_summary.py

@@ -3,7 +3,7 @@
 
 from temporalio import activity
 
-from ..clients.openai import client as openai_client
+from ..clients.model import julep_client
 from .types import RelationshipSummaryTaskArgs
 
 
@@ -49,7 +49,7 @@ async def run_prompt(
         )
     )
 
-    response = await openai_client.chat.completions.create(
+    response = await julep_client.chat.completions.create(
         model=model,
         messages=[
             {

agents-api/agents_api/activities/salient_questions.py

@@ -3,7 +3,7 @@
 
 from temporalio import activity
 
-from ..clients.openai import client as openai_client
+from ..clients.model import julep_client
 from .types import SalientQuestionsTaskArgs
 
 
@@ -40,7 +40,7 @@ async def run_prompt(
 ) -> str:
     prompt = make_prompt(SalientQuestionsTaskArgs(statements=statements, num=num))
 
-    response = await openai_client.chat.completions.create(
+    response = await julep_client.chat.completions.create(
         model=model,
         messages=[
             {

agents-api/agents_api/activities/summarization.py

@@ -10,7 +10,7 @@
     entries_summarization_query,
 )
 from agents_api.common.protocol.entries import Entry
-from agents_api.clients.openai import client as openai_client
+from agents_api.clients.model import julep_client
 
 
 example_previous_memory = """
@@ -130,7 +130,7 @@ async def run_prompt(
 ) -> str:
     prompt = make_prompt(dialog, previous_memories, **kwargs)
 
-    response = await openai_client.chat.completions.create(
+    response = await julep_client.chat.completions.create(
         model=model,
         messages=[
             {

agents-api/agents_api/autogen/__init__.py

@@ -1,3 +1,3 @@
 """
 This module contains automatically generated models based on the OpenAPI specification for the agents-api project. It includes definitions for key entities such as Users, Sessions, Agents, Tools, and their respective interactions. These models play a crucial role in defining the structure and constraints of data exchanged with the API endpoints, ensuring consistency and validation across the service. Generated models cover a wide range of functionalities from user management, session handling, agent configuration, to tool definitions, providing a comprehensive schema for the API's operations.
-"""
+"""

agents-api/agents_api/clients/__init__.py

@@ -6,4 +6,4 @@
 - `openai.py`: Facilitates interaction with OpenAI's API for natural language processing tasks.
 - `temporal.py`: Provides functionality for connecting to Temporal workflows, enabling asynchronous task execution and management.
 - `worker/__init__.py` and related files: Describe the role of the worker service client in sending tasks to be processed by an external worker service, focusing on memory management and other computational tasks.
-"""
+"""

agents-api/agents_api/clients/model.py

@@ -0,0 +1,10 @@
+from openai import AsyncOpenAI
+from ..env import model_inference_url, model_api_key, openai_api_key
+
+
+openai_client = AsyncOpenAI(api_key=openai_api_key)
+
+julep_client = AsyncOpenAI(
+    base_url=model_inference_url,
+    api_key=model_api_key,
+)

agents-api/agents_api/clients/worker/__init__.py

@@ -1,3 +1,3 @@
 """
 This module provides functionality for interacting with an external worker service. It includes utilities for creating and sending tasks, such as memory management tasks, to be processed by the service. The module leverages asynchronous HTTP requests via the `httpx` library to communicate with the worker service. Types for structuring task data are defined in `types.py`.
-"""
+"""

agents-api/agents_api/common/exceptions/__init__.py

@@ -8,6 +8,8 @@
 
 All custom exceptions extend from `BaseCommonException`, which encapsulates common attributes and behavior, including the error message and HTTP status code. This structured approach to exception handling facilitates precise and meaningful error feedback to API consumers, thereby improving the overall developer experience.
 """
+
+
 class BaseCommonException(Exception):
     def __init__(self, msg: str, http_code: int):
         super().__init__(msg)

agents-api/agents_api/common/exceptions/agents.py

@@ -2,6 +2,7 @@
 
 from uuid import UUID
 from . import BaseCommonException
+from agents_api.model_registry import ALL_AVAILABLE_MODELS
 
 
 class BaseAgentException(BaseCommonException):
@@ -39,3 +40,12 @@ def __init__(self, agent_id: UUID | str, doc_id: UUID | str):
         super().__init__(
             f"Doc {str(doc_id)} not found for agent {str(agent_id)}", http_code=404
         )
+
+
+class AgentModelNotValid(BaseAgentException):
+    def __init__(self, model: str):
+        super().__init__(
+            f"Unknown model: {model}. Please provide a valid model name."
+            "Known models are: " + ", ".join(ALL_AVAILABLE_MODELS.keys()),
+            http_code=400,
+        )

agents-api/agents_api/common/protocol/__init__.py

@@ -6,4 +6,4 @@
 - `SessionData`: Represents the data associated with a session, including identifiers and session-specific information such as situation, summary, and timestamps.
 
 These components are crucial for the effective operation and interaction within the agents API.
-"""
+"""

agents-api/agents_api/common/protocol/agents.py

@@ -1,5 +1,3 @@
-from typing import Literal
-
 from pydantic import BaseModel
 
 
@@ -19,9 +17,4 @@ class AgentDefaultSettings(BaseModel):
     """Penalty that decreases the likelihood of frequently used words in the agent's responses."""
     frequency_penalty: float = 0.0
     """Minimum probability threshold for including a word in the agent's response."""
-    min_p: float = 0.01
-
-
-ModelType = Literal[
-    "julep-ai/samantha-1", "julep-ai/samantha-1-turbo", "julep-ai/samantha-1-turbo-awq"
-]
+    min_p: float = 0.01