Initial commit: Dick Show Index pipeline with install script and docs

config.example.py

@@ -0,0 +1,118 @@
+# Dick Show Index - Configuration Template
+# Copy to config.py and customize
+
+import os
+
+# ============================================================================
+# Base Paths
+# ============================================================================
+BASE_DIR = os.path.dirname(os.path.abspath(__file__))
+
+# Data directories
+AUDIO_DIR = os.path.join(BASE_DIR, "audio")
+TRANSCRIPT_DIR = os.path.join(BASE_DIR, "transcripts")
+DATA_DIR = os.path.join(BASE_DIR, "data")
+DB_PATH = os.path.join(DATA_DIR, "index.db")
+EPISODES_JSONL = os.path.join(DATA_DIR, "episodes.jsonl")
+
+# RSS Feed
+FEED_URL = "https://thedickshow.com/feed/podcast/"
+
+# ============================================================================
+# Whisper Settings
+# ============================================================================
+# Model: tiny, base, small, medium, large-v3
+# smaller = faster, less accurate
+WHISPER_MODEL = os.environ.get("DS_MODEL", "small")
+
+# Device: cpu, cuda
+DEVICE = os.environ.get("DS_DEVICE", "cpu")
+
+# Compute type: int8, float16, float32
+# int8 = fastest, lowest memory (CPU)
+# float16 = GPU with tensor cores
+COMPUTE_TYPE = os.environ.get("DS_COMPUTE", "int8")
+
+# Language
+LANGUAGE = "en"
+
+# VAD filter removes silence/music - big speed win
+VAD_FILTER = True
+
+# Beam size (1 = greedy, 5 = better accuracy, slower)
+BEAM_SIZE = 1
+
+# Word timestamps (slower, more detailed)
+WORD_TIMESTAMPS = False
+
+# Condition on previous text (can propagate errors on long audio)
+CONDITION_ON_PREVIOUS_TEXT = False
+
+# ============================================================================
+# Audio Processing
+# ============================================================================
+TARGET_RATE = 16000  # Whisper expects 16kHz
+
+# HTTP settings
+DOWNLOAD_TIMEOUT = (15, 120)  # (connect, read) seconds
+CHUNK_SIZE = 1 << 20  # 1 MiB
+MAX_FEED_RETRIES = 3
+
+# Disk policy: delete MP3 after transcription (saves ~75GB for full backlog)
+KEEP_AUDIO = os.environ.get("DS_KEEP_AUDIO", "0") == "1"
+
+# ============================================================================
+# Analysis (opencode/LLM)
+# ============================================================================
+# Path to opencode binary
+OPENCODE = os.environ.get("OPENCODE_BIN", os.path.expanduser(r"~\.opencode\bin\opencode.exe"))
+
+# Model for analysis (see: opencode models)
+OPENCODE_MODEL = os.environ.get("OPENCODE_ANALYZE_MODEL", "opencode/nemotron-3-ultra-free")
+
+# Timeout for analysis (seconds)
+OPENCODE_TIMEOUT = int(os.environ.get("DS_ANALYZE_TIMEOUT", "600"))
+
+# Pause between analysis calls (seconds) - be nice to the provider
+ANALYSIS_PAUSE = int(os.environ.get("DS_ANALYZE_PAUSE", "45"))
+
+# ============================================================================
+# Memory / Rate Limit Guards
+# ============================================================================
+# Minimum free RAM before pausing transcription (bytes)
+MIN_FREE_RAM = int(os.environ.get("DS_MIN_FREE_RAM", str(1024 * 1024 * 1024)))  # 1 GB
+
+# Cooldown after rate limit hit (seconds)
+RATE_COOLDOWN_SECONDS = int(os.environ.get("DS_RATE_COOLDOWN", "900"))  # 15 min
+
+# ============================================================================
+# HTTP / Feed
+# ============================================================================
+DOWNLOAD_TIMEOUT = (15, 120)  # (connect, read)
+CHUNK_SIZE = 1 << 20  # 1 MiB
+MAX_FEED_RETRIES = 3
+
+# Ensure directories exist
+os.makedirs(AUDIO_DIR, exist_ok=True)
+os.makedirs(TRANSCRIPT_DIR, exist_ok=True)
+os.makedirs(os.path.join(os.path.dirname(__file__), "data"), exist_ok=True)
+
+# ============================================================================
+# Analysis Timeout (seconds)
+# ============================================================================
+ANALYSIS_TIMEOUT = int(os.environ.get("DS_ANALYZE_TIMEOUT", "600"))
+
+# Analysis pause between calls
+ANALYSIS_PAUSE = int(os.environ.get("DS_ANALYZE_PAUSE", "45"))
+
+# Minimum free RAM (bytes)
+MIN_FREE_RAM = int(os.environ.get("DS_MIN_FREE_RAM", str(1024 * 1024 * 1024)))
+
+# Rate cooldown seconds
+RATE_COOLDOWN_SECONDS = int(os.environ.get("DS_RATE_COOLDOWN", "900"))
+
+# ============================================================================
+# Directory Setup (auto-create)
+# ============================================================================
+for d in [AUDIO_DIR, TRANSCRIPT_DIR, os.path.join(os.path.dirname(__file__), "data")]:
+    os.makedirs(d, exist_ok=True)