LLMIFY THIS MOFO

internal/config/save.go

@@ -47,6 +47,9 @@ func Save(cfg *Config) error {
 		for name, pc := range cfg.Providers {
 			sb.WriteString(fmt.Sprintf("[providers.%s]\n", name))
 			sb.WriteString(fmt.Sprintf("  api_key = %q\n", pc.APIKey))
+			if pc.BaseURL != "" {
+				sb.WriteString(fmt.Sprintf("  base_url = %q\n", pc.BaseURL))
+			}
 			sb.WriteString("\n")
 		}
 	}
@@ -221,6 +224,9 @@ keywords = []
 #   api_key = ""               # ElevenLabs API key (or set ELEVENLABS_API_KEY env var)
 # [providers.deepgram]
 #   api_key = ""               # Deepgram API key (or set DEEPGRAM_API_KEY env var)
+# [providers.llama-swap]
+#   api_key = ""               # Or set LLAMA_SWAP_API_KEY
+#   base_url = "http://llama-swap.example:8080" # No /v1 suffix
 
 # ─────────────────────────────────────────────────────────────────────────────
 # Audio Recording
@@ -241,7 +247,7 @@ keywords = []
 # ─────────────────────────────────────────────────────────────────────────────
 
 [transcription]
-  provider = "openai"          # "openai", "groq-transcription", "mistral-transcription", "elevenlabs", "whisper-cpp"
+  provider = "openai"          # Also "llama-swap" for a remote OpenAI-compatible LlamaSwap server
   model = "whisper-1"          # Model: OpenAI="whisper-1", Groq="whisper-large-v3", Mistral="voxtral-mini-latest", ElevenLabs="scribe_v1"
   language = ""                # ISO 639-1 code (e.g., en, es, de). Empty for auto-detect.
   threads = 0                  # CPU threads for local transcription (0 = auto: uses NumCPU-1)
@@ -253,7 +259,7 @@ keywords = []
 
 [llm]
   enabled = true               # Enable LLM post-processing (highly recommended)
-  provider = "openai"          # "openai" or "groq" (must have API key configured above)
+  provider = "openai"          # "openai", "groq", or "llama-swap" (must have API key configured above)
   model = "gpt-4o-mini"        # OpenAI: "gpt-4o-mini", Groq: "llama-3.3-70b-versatile"
 
 [llm.post_processing]
@@ -272,7 +278,7 @@ keywords = []
 # ─────────────────────────────────────────────────────────────────────────────
 
 [injection]
-  backends = ["ydotool", "wtype", "clipboard"]  # Ordered fallback chain (tries each until one succeeds)
+  backends = ["clipboard-paste", "ydotool", "wtype", "clipboard"]  # Ordered fallback chain (tries each until one succeeds)
   ydotool_timeout = "5s"       # Timeout for ydotool commands
   wtype_timeout = "5s"         # Timeout for wtype commands
   clipboard_timeout = "3s"     # Timeout for clipboard operations
@@ -322,12 +328,15 @@ keywords = []
 # - "groq-transcription": Groq Whisper API (very fast, models: whisper-large-v3, whisper-large-v3-turbo)
 # - "mistral-transcription": Mistral Voxtral API (excellent for European languages, model: voxtral-mini-latest)
 # - "elevenlabs": ElevenLabs Scribe API (99 languages, models: scribe_v1, scribe_v2, scribe_v2_realtime)
+# - "llama-swap": Remote OpenAI-compatible LlamaSwap (any configured transcription model)
 #
 # LLM providers (for post-processing):
 # - "openai": GPT models (gpt-4o-mini recommended for cost/quality balance)
 # - "groq": Fast inference (llama-3.3-70b-versatile recommended)
+# - "llama-swap": Any chat model configured in your LlamaSwap server
 #
 # Injection backends:
+# - "clipboard-paste": Temporarily uses the regular clipboard + wtype Ctrl+V, then restores prior text clipboard.
 # - "ydotool": Uses ydotool (requires ydotoold daemon). Best for Chromium/Electron apps.
 # - "wtype": Uses wtype for Wayland. May have issues with some Chromium apps.
 # - "clipboard": Copies to clipboard only (most reliable, requires manual paste).