Enhance configuration management and expand documentation

.config/sample-config.yaml

@@ -0,0 +1,108 @@
+# For a detailed overview of each configuration option, see docs/configuration-options.md
+
+user:
+  name: Owl
+  client_token: <client-token>
+  voice_sample_filepath:
+
+web:
+  base_url: http://localhost
+  port: 3000
+  environment: development
+  api:
+    base_url: http://localhost
+    port: 8000
+
+llm:
+  model: ollama/mistral:instruct
+  base_url: http://localhost
+  port: 11434
+  api_key:
+
+async_transcription:
+  provider: deepgram
+
+streaming_transcription:
+  provider: deepgram
+
+deepgram:
+  api_key: <your-deepgram-api-key>
+  model: nova-2
+  language: en-US
+
+async_whisper:
+  host: localhost
+  port: 8010
+  hf_token: <your-hf-token>
+  device: cpu
+  compute_type: int8
+  batch_size: 16
+  model: tiny
+  verification_threshold: 0.1
+  verification_model_source: speechbrain/spkrec-ecapa-voxceleb
+  verification_model_directory: .models/spkrec-ecapa-voxceleb
+
+streaming_whisper:
+  host: localhost
+  port: 8009
+  model: small
+  language: en
+  silero_sensitivity: 0.4
+  webrtc_sensitivity: 2
+  post_speech_silence_duration: 0.5
+
+database:
+  url: sqlite:///./db.sqlite3
+
+captures:
+  directory: captures
+
+vad:
+  directory: .models/vad
+
+conversation_endpointing:
+  timeout_seconds: 300
+  min_utterances: 2
+
+notification:
+  apn_team_id:
+
+udp:
+  enabled: false
+  host: 0.0.0.0
+  port: 8001
+
+google_maps:
+  api_key:
+
+bing:
+  subscription_key:
+
+prompt:
+  suggest_links_system_message: >
+    You are the world's most advanced AI assistant. You are given the transcript of an interaction.
+    One of the participants is your client. Their name is {config.user.name}. Your task is to generate a
+    rich search query based on the summary of the interaction. You want to optimize the search query
+    to get maximally interesting relevant links for {config.user.name}. IMPORTANT: Try and make your search
+    query about a single subject that is most relevant to the interaction. Make it as specific as
+    possible and only pick one subject. Don't include {config.user.name}'s name in the output, just output
+    the query and nothing else. VERY IMPORTANT: You must just output the search engine query without
+    any prefix and nothing else!
+  summarization_system_message: >
+    You are the world's most advanced AI assistant. You are given the transcript of an interaction.
+    One of the participants is your client. Their name is {config.user.name}. The transcript includes
+    speaker ids, but unfortunately sometimes we don't know the specific person name and sometimes they
+    can be mislabeled. Do your best to infer the participants based on the context, but never referred
+    to the speaker ids in the summary because they alone are not useful. Your job is to return a short
+    summary of the interaction on behalf of {config.user.name} so they can remember what was happening.
+    This is for {config.user.name}'s memories so please include anything that might be useful but also
+    make it narrative so that it's helpful for creating a cherished memory. Format your summary with
+    the following sections: Summary, Atmosphere, Key Take aways (bullet points)
+  short_summarization_system_message: >
+    You are the world's most advanced AI assistant. You are given the transcript of an interaction. One
+    of the participants is your client. Their name is {config.user.name}. The transcript includes
+    speaker ids, but unfortunately sometimes we don't know the specific person name and sometimes they
+    can be mislabeled. Do your best to infer the participants based on the context, but never referred
+    to the speaker ids in the summary because they alone are not useful. Your job is to return a one
+    sentence summary of the interaction on behalf of {config.user.name}. It should capture the overall
+    significance of the interaction but not exceed one sentence.

.config/sample-env.env

@@ -0,0 +1,115 @@
+# For a detailed overview of each configuration option, see docs/configuration-options.md
+
+# User Configuration
+# ------------------
+# OWL_USER_CLIENT_TOKEN=
+# OWL_USER_NAME=
+# OWL_USER_VOICE_SAMPLE_FILEPATH=
+
+
+# Web Configuration
+# -----------------
+# OWL_WEB_BASE_URL=
+# OWL_WEB_PORT=
+# OWL_WEB_ENVIRONMENT=
+# OWL_WEB_BACKEND_BASE_URL=
+# OWL_WEB_BACKEND_PORT=
+
+
+# LLM Configuration
+# -----------------
+# OWL_LLM_MODEL=
+# OWL_LLM_BASE_URL=
+# OWL_LLM_PORT=
+# OWL_LLM_API_KEY=
+
+
+# Async Transcription Configuration
+# ---------------------------------
+# OWL_ASYNC_TRANSCRIPTION_PROVIDER=
+
+
+# Streaming Transcription Configuration
+# --------------------------------------
+# OWL_STREAMING_TRANSCRIPTION_PROVIDER=
+
+
+# Deepgram Configuration
+# ----------------------
+# OWL_DEEPGRAM_API_KEY=
+# OWL_DEEPGRAM_MODEL=
+# OWL_DEEPGRAM_LANGUAGE=
+
+
+# Async Whisper Configuration
+# ---------------------------
+# OWL_ASYNC_WHISPER_HOST=
+# OWL_ASYNC_WHISPER_PORT=
+# OWL_ASYNC_WHISPER_HF_TOKEN=
+# OWL_ASYNC_WHISPER_DEVICE=
+# OWL_ASYNC_WHISPER_COMPUTE_TYPE=
+# OWL_ASYNC_WHISPER_BATCH_SIZE=
+# OWL_ASYNC_WHISPER_MODEL=
+# OWL_ASYNC_WHISPER_VERIFICATION_THRESHOLD=
+# OWL_ASYNC_WHISPER_VERIFICATION_MODEL_SOURCE=
+# OWL_ASYNC_WHISPER_VERIFICATION_MODEL_DIRECTORY=
+
+
+# Streaming Whisper Configuration
+# -------------------------------
+# OWL_STREAMING_WHISPER_HOST=
+# OWL_STREAMING_WHISPER_PORT=
+# OWL_STREAMING_WHISPER_MODEL=
+# OWL_STREAMING_WHISPER_LANGUAGE=
+# OWL_STREAMING_WHISPER_SILERO_SENSITIVITY=
+# OWL_STREAMING_WHISPER_WEBRTC_SENSITIVITY=
+# OWL_STREAMING_WHISPER_POST_SPEECH_SILENCE_DURATION=
+
+
+# Database Configuration
+# ----------------------
+# OWL_DATABASE_URL=
+
+
+# Captures Configuration
+# ----------------------
+# OWL_CAPTURES_DIRECTORY=
+
+# VAD Configuration
+# -----------------
+# OWL_VAD_DIRECTORY=
+
+
+# Conversation Endpointing Configuration
+# --------------------------------------
+# OWL_CONVERSATION_ENDPOINTING_TIMEOUT_SECONDS=
+# OWL_CONVERSATION_ENDPOINTING_MIN_UTTERANCES=
+
+
+# Notification Configuration
+# --------------------------
+# OWL_NOTIFICATION_APN_TEAM_ID=
+
+
+# UDP Configuration
+# -----------------
+# OWL_UDP_ENABLED=
+# OWL_UDP_HOST=
+# OWL_UDP_PORT=
+
+
+# Google Maps Configuration
+# -------------------------
+#OWL_GOOGLE_MAPS_API_KEY=
+
+
+# Bing Configuration
+# ------------------
+# OWL_BING_SUBSCRIPTION_KEY=
+
+
+# Prompt Configuration
+# --------------------
+# OWL_PROMPT_SUGGEST_LINKS_SYSTEM_MESSAGE=
+# OWL_PROMPT_SUMMARIZATION_SYSTEM_MESSAGE=
+# OWL_PROMPT_SHORT_SUMMARIZATION_SYSTEM_MESSAGE=

.dockerignore

@@ -135,4 +135,4 @@ venv.bak/
 
 # Exclude the Dockerfile and .dockerignore to prevent potential leaks
 Dockerfile
-.dockerignore
+.dockerignore

.gitignore

@@ -1,5 +1,6 @@
 # Local configuration files
-/*.yaml
+.env
+config.yaml
 
 # Local database
 *.sqlite3
@@ -15,12 +16,11 @@ audio_cache/
 # vscode
 .vscode/launch.json
 
-
 # Local voice samples
 voice_samples/
 
 # Local models
-pretrained_models/
+.models/
 
 # Captures
 captures/
@@ -103,4 +103,4 @@ inject_derived
 .AppleDesktop
 Network Trash Folder
 Temporary Items
-.apdisk
+.apdisk