diff --git a/server-fastapi/README.md b/server-fastapi/README.md
deleted file mode 100644
index 194c2b3..0000000
--- a/server-fastapi/README.md
+++ /dev/null
@@ -1,256 +0,0 @@
-## ElatoAI: Realtime Voice AI Models on FastAPI
-
-`server-fastapi` is the simplest self-hosted Elato backend for people who want a normal Python server instead of an edge runtime.
-
-Use this if you want:
-
-- a FastAPI server you can run on your own machine or VM
-- a classic `STT -> LLM -> TTS` voice pipeline
-- a smaller provider surface that is easy to understand
-- the same ESP32 transport shape as the rest of Elato
-
-If you are new to the project, read these first:
-
-- `/Users/akashdeepdeb/Desktop/Projects/ElatoAI/README.md`
-- `/Users/akashdeepdeb/Desktop/Projects/ElatoAI/server/README.md`
-
-## The Simple Provider Set
-
-To keep onboarding straightforward, the classic FastAPI route is centered around a small set of providers.
-
-### LLM
-
-- `openai`
-- `claude`
-- `gemini`
-- `grok`
-
-### STT
-
-- `deepgram`
-- `whisper`
-
-### TTS
-
-- `elevenlabs`
-- `cartesia`
-- `deepgram`
-- `openai`
-
-The code still uses the `models/llm`, `models/stt`, and `models/tts` layout, but the active registry is intentionally trimmed so the default experience stays simple.
-
-## Default Setup
-
-The default classic route is:
-
-- STT: `deepgram`
-- LLM: `openai`
-- TTS: `elevenlabs`
-
-That gives people one obvious path to get running before they start swapping providers.
-
-## Project Layout
-
-```text
-server-fastapi/
-├── bot.py
-├── classic_route.py
-├── esp32_transport.py
-├── server.py
-├── env.example
-└── models/
-    ├── llm/
-    ├── stt/
-    └── tts/
-```
-
-## How The FastAPI Server Fits Into Elato
-
-Elato has three backend options right now:
-
-- `/Users/akashdeepdeb/Desktop/Projects/ElatoAI/server/deno`
-- `/Users/akashdeepdeb/Desktop/Projects/ElatoAI/server/cloudflare`
-- `/Users/akashdeepdeb/Desktop/Projects/ElatoAI/server-fastapi`
-
-A clean way to think about them is:
-
-- `Deno`: edge-first, mature provider integrations
-- `Cloudflare`: Workers + Durable Objects + Workers AI
-- `FastAPI`: normal Python server, easy to self-host, easy to reason about
-
-## Quick Start
-
-### 1. Create or activate your Python environment
-
-Use whatever you prefer. If you already use `uv`, that is a good default.
-
-### 2. Install dependencies
-
-This repo uses `pyproject.toml`, so install from that environment rather than a `requirements.txt` file.
-
-With `uv`:
-
-```bash
-cd /Users/akashdeepdeb/Desktop/Projects/ElatoAI/server-fastapi
-uv sync
-```
-
-Or with plain pip in your venv:
-
-```bash
-cd /Users/akashdeepdeb/Desktop/Projects/ElatoAI/server-fastapi
-pip install -e .
-```
-
-### 3. Create your env file
-
-Copy the example values from:
-
-- `/Users/akashdeepdeb/Desktop/Projects/ElatoAI/server-fastapi/env.example`
-
-Minimum example for the default route:
-
-```env
-DEEPGRAM_API_KEY=your_deepgram_api_key
-OPENAI_API_KEY=your_openai_api_key
-ELEVENLABS_API_KEY=your_elevenlabs_api_key
-
-CURRENT_VOICE_ROUTE=classic
-CLASSIC_STT_PROVIDER=deepgram
-CLASSIC_LLM_PROVIDER=openai
-CLASSIC_TTS_PROVIDER=elevenlabs
-
-ESP32_INPUT_SAMPLE_RATE=16000
-BROWSER_INPUT_SAMPLE_RATE=16000
-AUDIO_OUTPUT_SAMPLE_RATE=24000
-PIPELINE_AUDIO_IN_SAMPLE_RATE=16000
-PIPELINE_AUDIO_OUT_SAMPLE_RATE=24000
-
-ALLOWED_ORIGINS=*
-HOST=0.0.0.0
-PORT=7860
-```
-
-### 4. Run the server
-
-If you use `uv`:
-
-```bash
-cd /Users/akashdeepdeb/Desktop/Projects/ElatoAI/server-fastapi
-uv run server.py
-```
-
-If you use your activated venv directly:
-
-```bash
-cd /Users/akashdeepdeb/Desktop/Projects/ElatoAI/server-fastapi
-python server.py
-```
-
-### 5. Point your ESP32 at the FastAPI backend
-
-Update the firmware config so your hardware connects to this server instead of the Deno or Cloudflare backend.
-
-The ESP32 route is:
-
-```text
-/ws/esp32
-```
-
-For browser or Next.js testing, the server also exposes:
-
-- `/ws/browser`
-- `/ws/nextjs`
-
-## How Provider Selection Works
-
-The classic route reads three env vars:
-
-- `CLASSIC_STT_PROVIDER`
-- `CLASSIC_LLM_PROVIDER`
-- `CLASSIC_TTS_PROVIDER`
-
-So changing providers is just an env change.
-
-Examples:
-
-### OpenAI + Deepgram + ElevenLabs
-
-```env
-CLASSIC_STT_PROVIDER=deepgram
-CLASSIC_LLM_PROVIDER=openai
-CLASSIC_TTS_PROVIDER=elevenlabs
-```
-
-### Whisper + Claude + Cartesia
-
-```env
-CLASSIC_STT_PROVIDER=whisper
-CLASSIC_LLM_PROVIDER=claude
-CLASSIC_TTS_PROVIDER=cartesia
-```
-
-### Deepgram + Gemini + OpenAI TTS
-
-```env
-CLASSIC_STT_PROVIDER=deepgram
-CLASSIC_LLM_PROVIDER=gemini
-CLASSIC_TTS_PROVIDER=openai
-```
-
-## Unified Experience Across Elato
-
-A simple way to keep the product understandable is:
-
-- keep the Next.js frontend focused on character creation and device management
-- keep the ESP32 firmware focused on one transport protocol
-- let users choose one backend runtime:
-  - Deno
-  - Cloudflare
-  - FastAPI
-- inside each backend, expose the same conceptual knobs:
-  - `STT`
-  - `LLM`
-  - `TTS`
-
-That means the hardware story stays stable:
-
-- one firmware
-- one websocket-style mental model
-- three server deployment choices
-
-The cleanest unification strategy is not “every backend supports every provider.”
-It is:
-
-- every backend should expose the same categories
-- each backend should have one recommended default stack
-- advanced users can swap providers later
-
-## Recommended Defaults
-
-If you want a simple opinionated experience for users, keep one default combo per backend.
-
-Suggested defaults:
-
-- `Deno`: OpenAI realtime
-- `Cloudflare`: Workers AI STT/TTS + OpenAI LLM
-- `FastAPI`: Deepgram + OpenAI + ElevenLabs
-
-That gives users one obvious starting point without taking away flexibility.
-
-## Important Files
-
-If you want to change the FastAPI backend, start here:
-
-- `/Users/akashdeepdeb/Desktop/Projects/ElatoAI/server-fastapi/server.py`
-- `/Users/akashdeepdeb/Desktop/Projects/ElatoAI/server-fastapi/classic_route.py`
-- `/Users/akashdeepdeb/Desktop/Projects/ElatoAI/server-fastapi/esp32_transport.py`
-- `/Users/akashdeepdeb/Desktop/Projects/ElatoAI/server-fastapi/models/llm/__init__.py`
-- `/Users/akashdeepdeb/Desktop/Projects/ElatoAI/server-fastapi/models/stt/__init__.py`
-- `/Users/akashdeepdeb/Desktop/Projects/ElatoAI/server-fastapi/models/tts/__init__.py`
-
-## Current Notes
-
-- The filesystem still contains many scaffolded provider modules from the earlier broader experiment.
-- The active provider registry is now intentionally much smaller.
-- That means the codebase stays extensible, but the user-facing default path stays simple.
diff --git a/server-fastapi/__pycache__/classic_route.cpython-313.pyc b/server-fastapi/__pycache__/classic_route.cpython-313.pyc
deleted file mode 100644
index efd08aa..0000000
Binary files a/server-fastapi/__pycache__/classic_route.cpython-313.pyc and /dev/null differ
diff --git a/server-fastapi/classic_route.py b/server-fastapi/classic_route.py
deleted file mode 100644
index b8355a4..0000000
--- a/server-fastapi/classic_route.py
+++ /dev/null
@@ -1,55 +0,0 @@
-"""Classic STT -> LLM -> TTS pipeline builder."""
-
-from __future__ import annotations
-
-import os
-
-from character_prompt import LANGUAGE_LEARNING_PAL_PROMPT
-from loguru import logger
-from models.llm import create_llm_service
-from models.stt import create_stt_service
-from models.tts import create_tts_service
-from pipecat.audio.vad.silero import SileroVADAnalyzer
-from pipecat.audio.vad.vad_analyzer import VADParams
-from pipecat.processors.aggregators.llm_context import LLMContext
-from pipecat.processors.aggregators.llm_response_universal import (
-    LLMContextAggregatorPair,
-    LLMUserAggregatorParams,
-)
-
-
-def build_classic_route(input_processor, context: LLMContext):
-    stt_provider = os.getenv("CLASSIC_STT_PROVIDER", "deepgram")
-    llm_provider = os.getenv("CLASSIC_LLM_PROVIDER", "openai")
-    tts_provider = os.getenv("CLASSIC_TTS_PROVIDER", "elevenlabs")
-
-    logger.info(
-        "Building classic route with stt={} llm={} tts={}",
-        stt_provider,
-        llm_provider,
-        tts_provider,
-    )
-
-    stt = create_stt_service(stt_provider)
-    llm = create_llm_service(
-        llm_provider,
-        system_instruction=LANGUAGE_LEARNING_PAL_PROMPT,
-    )
-    tts = create_tts_service(tts_provider)
-
-    user_aggregator, assistant_aggregator = LLMContextAggregatorPair(
-        context,
-        user_params=LLMUserAggregatorParams(
-            vad_analyzer=SileroVADAnalyzer(params=VADParams(stop_secs=1))
-        ),
-    )
-
-    processors = [
-        input_processor,
-        stt,
-        user_aggregator,
-        llm,
-        tts,
-    ]
-
-    return processors, assistant_aggregator
diff --git a/server-fastapi/env.example b/server-fastapi/env.example
deleted file mode 100644
index 8f22444..0000000
--- a/server-fastapi/env.example
+++ /dev/null
@@ -1,27 +0,0 @@
-DEEPGRAM_API_KEY=your_deepgram_api_key
-OPENAI_API_KEY=your_openai_api_key
-ANTHROPIC_API_KEY=your_anthropic_api_key
-GEMINI_API_KEY=your_gemini_api_key
-XAI_API_KEY=your_xai_api_key
-ELEVENLABS_API_KEY=your_elevenlabs_api_key
-CARTESIA_API_KEY=your_cartesia_api_key
-
-# Classic route providers
-CURRENT_VOICE_ROUTE=classic
-CLASSIC_STT_PROVIDER=deepgram
-CLASSIC_LLM_PROVIDER=openai
-CLASSIC_TTS_PROVIDER=elevenlabs
-
-# Transport and pipeline sample rates
-ESP32_INPUT_SAMPLE_RATE=16000
-BROWSER_INPUT_SAMPLE_RATE=16000
-AUDIO_OUTPUT_SAMPLE_RATE=24000
-PIPELINE_AUDIO_IN_SAMPLE_RATE=16000
-PIPELINE_AUDIO_OUT_SAMPLE_RATE=24000
-
-# Browser / Next.js access
-ALLOWED_ORIGINS=*
-
-# WebSocket server settings
-HOST=0.0.0.0
-PORT=7860
diff --git a/server-fastapi/models/llm/__init__.py b/server-fastapi/models/llm/__init__.py
deleted file mode 100644
index fe825f2..0000000
--- a/server-fastapi/models/llm/__init__.py
+++ /dev/null
@@ -1,20 +0,0 @@
-"""LLM provider registry."""
-
-from __future__ import annotations
-
-from models._provider_loader import load_provider_factory
-
-LLM_REGISTRY = {
-    "claude": "models.llm.anthropic",
-    "anthropic": "models.llm.anthropic",
-    "gemini": "models.llm.google_gemini",
-    "google_gemini": "models.llm.google_gemini",
-    "google_vertex_ai": "models.llm.google_vertex_ai",
-    "grok": "models.llm.grok",
-    "openai": "models.llm.openai",
-}
-
-
-def create_llm_service(provider_name: str, **kwargs):
-    factory = load_provider_factory(LLM_REGISTRY, provider_name, "LLM")
-    return factory(**kwargs)
diff --git a/server-fastapi/models/llm/__pycache__/__init__.cpython-313.pyc b/server-fastapi/models/llm/__pycache__/__init__.cpython-313.pyc
deleted file mode 100644
index 26d562f..0000000
Binary files a/server-fastapi/models/llm/__pycache__/__init__.cpython-313.pyc and /dev/null differ
diff --git a/server-fastapi/models/stt/__init__.py b/server-fastapi/models/stt/__init__.py
deleted file mode 100644
index 26cb9d7..0000000
--- a/server-fastapi/models/stt/__init__.py
+++ /dev/null
@@ -1,16 +0,0 @@
-"""STT provider registry."""
-
-from __future__ import annotations
-
-from models._provider_loader import load_provider_factory
-
-STT_REGISTRY = {
-    "deepgram": "models.stt.deepgram",
-    "openai": "models.stt.openai",
-    "whisper": "models.stt.whisper",
-}
-
-
-def create_stt_service(provider_name: str, **kwargs):
-    factory = load_provider_factory(STT_REGISTRY, provider_name, "STT")
-    return factory(**kwargs)
diff --git a/server-fastapi/models/stt/__pycache__/__init__.cpython-313.pyc b/server-fastapi/models/stt/__pycache__/__init__.cpython-313.pyc
deleted file mode 100644
index 4aaac67..0000000
Binary files a/server-fastapi/models/stt/__pycache__/__init__.cpython-313.pyc and /dev/null differ
diff --git a/server-fastapi/models/tts/__init__.py b/server-fastapi/models/tts/__init__.py
deleted file mode 100644
index efc7468..0000000
--- a/server-fastapi/models/tts/__init__.py
+++ /dev/null
@@ -1,17 +0,0 @@
-"""TTS provider registry."""
-
-from __future__ import annotations
-
-from models._provider_loader import load_provider_factory
-
-TTS_REGISTRY = {
-    "cartesia": "models.tts.cartesia",
-    "deepgram": "models.tts.deepgram",
-    "elevenlabs": "models.tts.elevenlabs",
-    "openai": "models.tts.openai",
-}
-
-
-def create_tts_service(provider_name: str, **kwargs):
-    factory = load_provider_factory(TTS_REGISTRY, provider_name, "TTS")
-    return factory(**kwargs)
diff --git a/server-fastapi/models/tts/__pycache__/__init__.cpython-313.pyc b/server-fastapi/models/tts/__pycache__/__init__.cpython-313.pyc
deleted file mode 100644
index eaf3005..0000000
Binary files a/server-fastapi/models/tts/__pycache__/__init__.cpython-313.pyc and /dev/null differ
diff --git a/server/fastapi/README.md b/server/fastapi/README.md
index d11ee4d..e443b67 100644
--- a/server/fastapi/README.md
+++ b/server/fastapi/README.md
@@ -54,7 +54,7 @@ That gives people one obvious path to get running before they start swapping pro
 ```text
 server/fastapi/
 ├── bot.py
-├── classic_route.py
+├── voice_pipeline.py
 ├── esp32_transport.py
 ├── server.py
 ├── env.example
@@ -327,7 +327,7 @@ That gives users one obvious starting point without taking away flexibility.
 If you want to change the FastAPI backend, start here:
 
 - `/Users/akashdeepdeb/Desktop/Projects/ElatoAI/server/fastapi/server.py`
-- `/Users/akashdeepdeb/Desktop/Projects/ElatoAI/server/fastapi/classic_route.py`
+- `/Users/akashdeepdeb/Desktop/Projects/ElatoAI/server/fastapi/voice_pipeline.py`
 - `/Users/akashdeepdeb/Desktop/Projects/ElatoAI/server/fastapi/esp32_transport.py`
 - `/Users/akashdeepdeb/Desktop/Projects/ElatoAI/server/fastapi/models/llm/__init__.py`
 - `/Users/akashdeepdeb/Desktop/Projects/ElatoAI/server/fastapi/models/stt/__init__.py`
diff --git a/server/fastapi/bot.py b/server/fastapi/bot.py
index b6c4207..e766c3a 100644
--- a/server/fastapi/bot.py
+++ b/server/fastapi/bot.py
@@ -9,7 +9,7 @@
 import os
 from typing import Literal
 
-from classic_route import build_classic_route
+from voice_pipeline import build_voice_pipeline
 from dotenv import load_dotenv
 from gem_live_route import build_gem_live_route
 from grok_route import build_grok_route
@@ -155,7 +155,7 @@ async def run_bot_session(
     elif voice_route == "grok":
         route_processors, assistant_aggregator = build_grok_route(input_processor, context)
     else:
-        route_processors, assistant_aggregator = build_classic_route(input_processor, context)
+        route_processors, assistant_aggregator = build_voice_pipeline(input_processor, context)
 
     processors = [transport.input(), *route_processors]
 
diff --git a/server/fastapi/classic_route.py b/server/fastapi/voice_pipeline.py
similarity index 92%
rename from server/fastapi/classic_route.py
rename to server/fastapi/voice_pipeline.py
index b8355a4..8fa61de 100644
--- a/server/fastapi/classic_route.py
+++ b/server/fastapi/voice_pipeline.py
@@ -1,4 +1,4 @@
-"""Classic STT -> LLM -> TTS pipeline builder."""
+"""Default STT -> LLM -> TTS voice pipeline builder."""
 
 from __future__ import annotations
 
@@ -18,7 +18,7 @@ from pipecat.processors.aggregators.llm_response_universal import (
 )
 
 
-def build_classic_route(input_processor, context: LLMContext):
+def build_voice_pipeline(input_processor, context: LLMContext):
     stt_provider = os.getenv("CLASSIC_STT_PROVIDER", "deepgram")
     llm_provider = os.getenv("CLASSIC_LLM_PROVIDER", "openai")
     tts_provider = os.getenv("CLASSIC_TTS_PROVIDER", "elevenlabs")