projectdavid 1.41.4

Python SDK for interacting with the Entities Assistant API.

Entity — by Project David

The Entity SDK is a composable, Pythonic interface to the Entities API for building intelligent applications across local, open-source, and cloud LLMs.

It unifies:

• Users, threads, assistants, messages, runs, inference
• Function calling, code interpretation, and structured streaming
• Vector memory, file uploads, and secure tool orchestration

Local inference is fully supported via Ollama.

---

🔌 Supported Inference Providers

Provider	Type
Ollama	Local (Self-Hosted)
DeepSeek	☁ Cloud (Open-Source)
Hyperbolic	☁ Cloud (Proprietary)
OpenAI	☁ Cloud (Proprietary)
Together AI	☁ Cloud (Aggregated)
Azure Foundry	☁ Cloud (Enterprise)

---

📦 Installation

pip install projectdavid

---

Quick Start

import os

from dotenv import load_dotenv
from projectdavid import Entity

load_dotenv()

# --------------------------------------------------
# Load the Entities client with your user API key
# Note: if you define ENTITIES_API_KEY="ea_6zZiZ..."
# in .env, you do not need to pass in the API key directly.
# We pass in here directly for clarity
# ---------------------------------------------------
client = Entity(base_url="http://localhost:9000", api_key=os.getenv("ENTITIES_API_KEY"))

user_id = "user_kUKV8octgG2aMc7kxAcD3i"

# -----------------------------
# create an assistant
# ------------------------------
assistant = client.assistants.create_assistant(
    name="test_assistant",
    instructions="You are a helpful AI assistant",
)
print(f"created assistant with ID: {assistant.id}")

# -----------------------------------------------
# Create a thread
# Note:
# - Threads are re-usable
# Reuse threads in the case you want as continued
# multi turn conversation
# ------------------------------------------------
print("Creating thread...")
thread = client.threads.create_thread(participant_ids=[user_id])

print(f"created thread with ID: {thread.id}")
# Store the dynamically created thread ID
actual_thread_id = thread.id


# -----------------------------------------
#  Create a message using the NEW thread ID
# --------------------------------------------
print(f"Creating message in thread {actual_thread_id}...")
message = client.messages.create_message(
    thread_id=actual_thread_id,
    role="user",
    content="Hello, assistant! Tell me about the latest trends in AI.",
    assistant_id=assistant.id,
)
print(f"Created message with ID: {message.id}")

# ---------------------------------------------
# step 3 - Create a run using the NEW thread ID
# ----------------------------------------------
print(f"Creating run in thread {actual_thread_id}...")
run = client.runs.create_run(assistant_id=assistant.id, thread_id=actual_thread_id)
print(f"Created run with ID: {run.id}")

# ------------------------------------------------
# Instantiate the synchronous streaming helper
# --------------------------------------------------
sync_stream = client.synchronous_inference_stream

# ------------------------------------------------------
# step 4 - Set up the stream using the NEW thread ID
# --------------------------------------------------------
print(f"Setting up stream for thread {actual_thread_id}...")
sync_stream.setup(
    user_id=user_id,
    thread_id=actual_thread_id,
    assistant_id=assistant.id,
    message_id=message.id,
    run_id=run.id,
    api_key=os.getenv("YOUR_API_KEY"),
)
print("Stream setup complete. Starting streaming...")

# --- Stream initial LLM response ---
try:
    for chunk in sync_stream.stream_chunks(
        provider="Hyperbolic",
        model="hyperbolic/deepseek-ai/DeepSeek-V3-0324",  # Ensure this model is valid/available
        timeout_per_chunk=15.0,
    ):
        content = chunk.get("content", "")
        if content:
            print(content, end="", flush=True)
    print("\n--- End of Stream ---")  # Add newline after stream
except Exception as e:
    print(f"\n--- Stream Error: {e} ---")  # Catch errors during streaming

print("Script finished.")

Model Routes

The script above maps each model to a route suffix that you use when calling the API. For example, to invoke the DeepSeek V3 model hosted on Hyperbolic you would use the suffix:

hyperbolic/deepseek-ai/DeepSeek-V3-0324

Below is a table that lists the route suffix for every supported model.

Below is a table that lists the route suffix for every supported model.

View Model Routes Table

The assisants response:

Hello! The field of AI is evolving rapidly, and here are some of the latest trends as of early 2025:

1. Multimodal AI Models

• Models like GPT-4, Gemini, and others now seamlessly process text, images, audio, and video in a unified way, enabling richer interactions (e.g., ChatGPT with vision).
• Applications include real-time translation with context, AI-generated video synthesis, and more immersive virtual assistants.

2. Smaller, More Efficient Models

• While giant models (e.g., GPT-4, Claude 3) still dominate, there’s a push for smaller, specialized models (e.g., Microsoft’s Phi-3, Mistral 7B) that run locally on devices with near-LLM performance.
• Focus on energy efficiency and reduced computational costs.

3. AI Agents & Autonomous Systems

• AI “agents” (e.g., OpenAI’s “Agentic workflows”) can now perform multi-step tasks autonomously, like coding, research, or booking trips.
• Companies are integrating agentic AI into workflows (e.g., Salesforce, Notion AI).

4. Generative AI Advancements

• Video generation: Tools like OpenAI’s Sora, Runway ML, and Pika Labs produce high-quality, longer AI-generated videos.
• 3D asset creation: AI can now generate 3D models from text prompts (e.g., Nvidia’s tools).
• Voice cloning: Ultra-realistic voice synthesis (e.g., ElevenLabs) is raising ethical debates.

5. Regulation & Ethical AI

• Governments are catching up with laws like the EU AI Act and U.S. executive orders on AI safety.
• Watermarking AI content (e.g., C2PA standards) is gaining traction to combat deepfakes.

6. AI in Science & Healthcare

• AlphaFold 3 (DeepMind) predicts protein interactions with unprecedented accuracy.
• AI-driven drug discovery (e.g., Insilico Medicine) is accelerating clinical trials.

7. Open-Source vs. Closed AI

• Tension between open-source (Mistral, Meta’s Llama 3) and proprietary models (GPT-4, Gemini) continues, with debates over safety and innovation.

8. AI Hardware Innovations

• New chips (e.g., Nvidia’s Blackwell, Groq’s LPUs) are optimizing speed and cost for AI workloads.
• “AI PCs” with NPUs (neural processing units) are becoming mainstream.

9. Personalized AI

• Tailored AI assistants learn individual preferences (e.g., Rabbit R1, Humane AI Pin).
• Privacy-focused local AI (e.g., Apple’s on-device AI in iOS 18).

10. Quantum AI (Early Stages)

• Companies like Google and IBM are exploring quantum machine learning, though practical applications remain limited.

Would you like a deeper dive into any of these trends?

---

Documentation

Domain	Link
Assistants	assistants.md
Threads	threads.md
Messages	messages.md
Runs	runs.md
Inference	inference.md
Streaming	streams.md
Tools	function_calls.md
Code Interpretation	code_interpretation.md
Files	files.md
Vector Store(RAG)	vector_store.md
Versioning	versioning.md

---

✅ Compatibility & Requirements

• Python 3.10+
• Compatible with local or cloud deployments of the Entities API

---

Related Repositories

• Entities API — containerized API backend
• entities_common — shared validation, schemas, utilities, and tools. This package is auto installed as dependency of entities SDK or entities API.