This implements the minimal ol invoke Professor Caraza-Harter requested:

• POST-only, JSON-only

• Usage:
  ol invoke myfunc # POST null body
  ol invoke myfunc '{"A":"B"}' # POST JSON body
  ol invoke myfunc '{"A":"B"}' -p default-ol # project-aware

• Resolves URL like ol worker up via deploy dir (default project default-ol).
  For container-based dev on macOS, it also respects $OL_URL if set (so
  OL_URL=http://host.docker.internal:5000 works from inside Docker).

• Flags:
  -p, --project (project/deploy name)
  --pretty (pretty-print JSON)
  --timeout N (seconds)

Manual test evidence:

• With mock worker on host port 5000:
  ./ol-linux invoke -p default-ol --pretty myfunc
  ./ol-linux invoke -p default-ol --pretty myfunc '{"A":"B"}'
  Both returned 200 with expected echo payloads.

Follow-up:

• I’ll test on Ubuntu 24.04 with a real worker (ol worker up -d -p default-ol)
  and report results on this PR thread.

Update: Ubuntu 24.04 Verification Results 🟢

I have successfully verified the implementation on a fresh Ubuntu 24.04 VM with a real OpenLambda worker instance.

Test Environment:

• OS: Ubuntu 24.04 LTS
• Build: Built binary from source (go build .)
• Worker: Real worker initialized on port 5000 (ol worker up -p test-worker)

Test Cases Executed:

1. Project-Aware Lookup:

   • Command: ./ol-linux invoke -p test-worker hello-lambda-example
   • Result: Success ("hello")
   • Confirmed that the CLI correctly resolved the port from the custom project directory.

2. JSON Payload (Cmdline):

   • Command: ./ol-linux invoke -p test-worker hello-lambda-example '{"some":"json"}'
   • Result: Success ("hello")
   • Confirmed that JSON strings are accepted and transmitted correctly.

3. Input Validation (Negative Test):

   • Command: ./ol-linux invoke ... '{"broken":"json"'
   • Result: Error: invalid JSON
   • Confirmed that the CLI validates input before transmission.

4. End-to-End Latency:

   • Verified via worker.out logs that the request triggered a sandbox unpause event.

Ready for review!