Invoke agent flows from streamlit and apex app.

[jrgauthier01]

Description

two code samples to invoke agent flows from streamlit and apex. Those have been written by our engineering team.

[ahmedawan-oracle]

PR Review

Solid sample code — the APEX plugin architecture (AQ-backed async processing, ownership validation, apex_escape.html() throughout) is production-quality. A few items to fix before merge:

Must Fix

1. ~15 DEBUG: print() statements left in aidp_chat.py

The StreamingResponse.__iter__ method is full of lines like:

print(f"DEBUG: Raw line: {decoded_line[:200]}")
print(f"DEBUG: Parsed chunk keys: {list(chunk_json.keys())}")
print(f"DEBUG: ✓ Yielding text chunk (length={len(text)})")

These will spam server logs. Remove them or gate behind logging.DEBUG.

2. Init Client button is broken (aidp_streamlit_chat.py ~line 862)

if init_clicked:
    st.session_state.messages = [...]
    st.session_state.last_response_id = None
    st.session_state.trace_history = []
    st.rerun()          # <-- halts execution here
    ensure_client_ready()  # <-- unreachable

st.rerun() stops execution immediately, so ensure_client_ready() never runs. Move ensure_client_ready() before the st.rerun() call.

3. Refresh button is a no-op (aidp_streamlit_chat.py ~line 871)

if refresh_clicked:
    print("Refresh")

Either implement it or remove the button.

4. Placeholder URL never updated (aidp_streamlit_chat.py ~line 720)

📚 See [GETTING_STARTED.md](https://github.com/your-repo/blob/main/GETTING_STARTED.md) for detailed examples!

This is a literal your-repo placeholder — should point to the actual repo path.

Minor

• aidp_chat.py: os imported twice (lines 1 & 8), List, Dict, Any imported twice (lines 3 & 5)
• aidp_chat.py _post(): mutable default argument headers={"Content-Type":"application/json"} — use headers=None with assignment inside
• requirements.txt: streamlit-authenticator>=0.3.2 is listed but never imported anywhere
• create_credential_and_profile.sql: no idempotency guard — re-running fails if ACP_CHAT_CRED already exists
• summarize_conversation() AI prompt: typo "acronymm" (double m)