Proscenium: Framing Agents as Actors

Proscenium establishes a modular analysis of AI "Agents". It is composed of several well-defined subsystems.

A few of the initial goals of this project include:

• Clarify how the components interact
• Identify areas where innovation is still redefining interfaces
• Highlight designs that can limit the "blast radius" of changes
• For users of frameworks, identify risk of lock-in
• Enumerate "glue code", libraries, or protocols that are missing from the ecosystem

Subsystems

+------------------------------------------------------------------------------+
|                                     Scripts                                  |
+------------------------------------------------------------------------------+
|                                     Actors                                   |
+------------+--+------------+--+--------------+--+-------+--+-------+--+------+
|   Invoke   |  |  Complete  |  |   Remember   |  | Chunk |  | Parse |  | Load |
+------------+  +------------+  +--------------+  +-------+  +-------+  +------+
| APIs,      |  |  Inference |  | DB, RAM, ... |  | ...   |  | ...   |  | ...  |
| Libraries, |  |  Providers |  +--------------+  +-------+  +-------+  +------+
| ...        |  +------------+
+------------+  |    LLMs    |
                +------------+

Methodology

The purpose of this repository is to show several examples of these sub-systems working together in a way that mimics patterns highlighted in modern agentic frameworks.

The will include

• Tool use
• Reflection
• RAG
• GraphRAG
• ...

Subsystem implementations used in these demos:

• Actors
  • Thespian
• Inference
  • AI Suite
• Functions
  • Gofannon
• Memory
  • RAM: local module proscenium.memory

Setup

git clone [email protected]:The-AI-Alliance/proscenium.git

cd proscenium

python -m venv venv

. venv/bin/activate

python -m pip install .

python -m example.rag.build_vector_db

or

git clone [email protected]:The-AI-Alliance/proscenium.git

cd proscenium

docker compose -f .devcontainer/docker-compose.yml up

docker exec -it devcontainer-devcontainer-1 sh -c "cd workspaces/proscenium && python -m example.rag.build_vector_db"

or