feat: stub out voice control integration, gcloud not yet work ing

.gitignore

@@ -178,3 +178,4 @@ session-log
 dist
 sample-data
 scratch.md
+voice-tmp.wav

.husky/pre-commit

@@ -1,3 +1,3 @@
 #!/bin/sh
 
-bun run check-for-api-key
+bun run ./scripts/check-for-api-key

README.md

@@ -26,6 +26,7 @@ A very simple Typescript framework to build tool-wielding AI Agents.
 - A few fun pre-baked Agents
 - Use Agents in either CLI and Server Modes
 - Prompt caching support
+- Beta: Voice mode (speak to the agent)
 
 ## Setup
 
@@ -314,4 +315,24 @@ Agent.create({
 });
 ```
 
-This setup will cache both the system prompt and tools, in that order, up to the Anthropic-imposed limit.
+### Voice Mode
+
+Voice commands are available when using the `EasyAgentCLI`. To enable voice mode:
+
+As CLI mode:
+
+Start CLI
+```bash
+bun run cli --voice
+```
+
+As library:
+
+```ts
+import { EasyAgentCLI } from "easy-agent";
+import MyAgent from "./some-location-where-i-defined-it";
+
+EasyAgentCLI.start([MyAgent], { voice: true });
+```
+
+Voice mode uses Google Speech to Text, so you'll need valid Google Cloud API credentials in your environment to use it. You'll also need to enable the Speech to Text API in your Google Cloud Console.

TODO

@@ -1,15 +1,19 @@
 ## Todo
+- [] Make sure you can pause recording so you can read to figure out the next thing you're gonna say.
+    - [] press enter to continue
+- [] The keyboard handlers inside the recording module are flaky. They appear to register more than once and it makes it impossible to kill the process and duplicates logs.
 
-GOAL
-- Take this framework live to NPM
-   - take ai project manager live (based on published NPM package)
-- Adding Gemini support
-    - Gemini has a couple things that it seems better at.
+- ideas
+    - figure out why google speech to text doesn't work.. not even their boilerplate one
+    - do VOSK
+    - do DeepSpeech
 
-Next:
-- bundle as proper library
-- make library example
+- [] Voice support
+    - [] README update with TOC
+    - [] README update with voice
 
+- [] Add Gemini support
+    - Gemini has a couple things that it seems better at.
 
 Future:
 - refactor out rendering logic from client to a CLI-oriented module, so that I can break out the message lifecycle from the rendering for server use