api_server.py

# api_server.py

import requests
from fastapi import FastAPI, Request
from pydantic import BaseModel
from fastapi.middleware.cors import CORSMiddleware
from fastapi.responses import HTMLResponse
from fastapi.staticfiles import StaticFiles
from fastapi.responses import FileResponse
from fastapi import BackgroundTasks

from core.conversation_manager import ConversationState
from core.conversation_engine import handle_turn

from core.agent_orchestrator import run_agents
from core.input_classifier import is_meaningful_input
from core.memory_store import save_state, load_state # session persistence handling
from core.conversation_manager import from_dict # to have the retrieved conversation in DICT/JSON converted to ConversationState Object
from core.command_parser import parse_command

from agents.epistemic_agent import epistemic_response

from core.epistemic.profile_manager import update_global_profile
from core.epistemic.epistemic_cross import cross_session_analysis

# CREATE APP FIRST
app = FastAPI()

# THEN MOUNT STATIC
app.mount("/static", StaticFiles(directory="web"), name="static")

# sessions = {}

# ---- CORS FIX ----
app.add_middleware(
    CORSMiddleware,
    allow_origins=["*"],  # allow any origin
    allow_credentials=True,
    allow_methods=["*"],
    allow_headers=["*"],
)

class ChatRequest(BaseModel):
    session_id: str
    message: str


@app.get("/", response_class=HTMLResponse)
def serve_ui():
    return FileResponse("web/index.html")

#@app.get("/")
#def root():
#    return {"status": "CGJI01 API running"}

#@app.post("/analyze")
#def analyze(req: Request):
#    results = run_agents(req.text)
#    return results


@app.post("/analyze")
def analyze(req: Request):

    if not is_meaningful_input(req.text):
        return {
            "status": "ok",
            "message": "Input does not appear to require deep psychological analysis.",
            "input": req.text
        }

    results = run_agents(req.text)

    return {"result": results}
    
"""

sessions = {}

    

@app.post("/chat")

async def chat(req: Request):

    data = await req.json()

    session_id = data["session_id"]

    message = data["message"]



    if session_id not in sessions:

        sessions[session_id] = ConversationState()



    state = sessions[session_id]



    output = handle_turn(state, message)



    if output["type"] == "end":

        del sessions[session_id]  # Clean up session when finished



    return output

"""    

# HANDLE AGENT SWITCH - FUNCTION
def handle_agent_switch(state, agent_name):

    output = run_specific_agent(agent_name, state.initial_query)

    # Ensure output is a dict
    if not isinstance(output, dict):
        output = {
            "type": "error",
            "content": "Agent returned invalid response"
        }
    
    print("🧠 AGENT SWITCH OUTPUT:", output)
    
    # Inject agent identity
    output["agent"] = agent_name

    return output



#def handle_agent_switch(state, agent_name):

    # 🔥 TEMP: route through existing system
#    print(f"⚠️ Switching to {agent_name}, but using handle_turn for now")

#    return handle_turn(state, state.initial_query)

def run_epistemic_witness(state, session_id):
    try:
        # Extract full histories
        client_history = [
            msg["text"] for msg in state.agent_histories.get("jung", [])
            if msg["role"] == "client"
        ]

        jung_history = [
            msg["text"] for msg in state.agent_histories.get("jung", [])
            if msg["role"] == "agent"
        ]

        # Run epistemic analysis
        analysis = epistemic_response(client_history, jung_history)

        # Store inside state
        if not hasattr(state, "witness_logs"):
            state.witness_logs = {}

        state.witness_logs["epistemic"] = analysis

        print("🧠 EPISTEMIC UPDATED")

    except Exception as e:
        print("❌ Epistemic error:", e)


sessions = {}

@app.post("/chat")
def chat(req: ChatRequest, background_tasks: BackgroundTasks):

    print("🔥 CHAT ENDPOINT HIT")

    session_id = req.session_id
    message = req.message

    print("\n🧾 RAW REQUEST BODY:", req)

    #if session_id not in sessions:
    #    sessions[session_id] = ConversationState()

    if session_id not in sessions:
        saved_data = load_state(session_id) # when user sends a message after loading, backend restores full conversation automatically

        if saved_data:
            print("♻️ Restored session from disk")
            #sessions[session_id] = loaded

            print("🔍 SAVED DATA:", saved_data)  
            try:
                state = from_dict(saved_data)
            except Exception as e:
                print("❌ RESTORE FAILED:", e)
                state = ConversationState()

        else:
            #sessions[session_id] = ConversationState()
            state = ConversationState()

        sessions[session_id] = state

    state = sessions[session_id]

    # ✅ Ensure flag exists
    if not hasattr(state, "conversation_started"):
        state.conversation_started = False

    if not state.conversation_started:
        state.conversation_started = True

    # ✅ Store initial query
    if not state.initial_query and not message.startswith("__agent__:"):
        #state.initial_query = message
        if state.history:
            state.initial_query = next(
            (m["content"] for m in data.get("history", []) if m["role"] == "client"),
            ""
            ) # improve initial query restore
        else:
            state.initial_query = message

    print("\n📥 USER MESSAGE:", message)


    cmd, cleaned_message = parse_command(message)

    if cmd:
        if cmd["command"] == "general":
            state.mode = "general"
            state.role = "client"

        elif cmd["command"] == "analysis":
            state.mode = "analysis"

            if len(cmd["args"]) > 0:
                if cmd["args"][0] == "analyst":
                    state.role = "analyst"
                elif cmd["args"][0] == "client":
                    state.role = "client"

        message = cleaned_message



    # -----------------------------
    # AGENT SWITCH
    # -----------------------------
    if message.startswith("__agent__:"):
        selected_agent = message.split(":")[1]

        print("🧭 Switching to agent:", selected_agent)

        state.current_agent = selected_agent

        output = handle_agent_switch(state, selected_agent)

        print("\n📤 RESPONSE TO UI:", output)
        return output

    # -----------------------------
    # NORMAL FLOW (JUNG etc.)
    # -----------------------------
    output = handle_turn(state, message)

     
    # ---- (EXPERIMENTAL) LOW-LEVEL CROSS-SESSION EPISTEMIC INTELLIGENCE/ANALYSIS ----
    
    # -------------------------------
    # UPDATE GLOBAL PROFILE
    # -------------------------------
    state.session_id = session_id  # ensure attached
    session_record = update_global_profile(state)

    # -------------------------------
    # EPISTEMIC CROSS-SESSION
    # -------------------------------
    epistemic_text = cross_session_analysis(session_record)

    # attach to response
    output["witness_updates"] = {
        "epistemic": epistemic_text
    }
    # ---- (EXPERIMENTAL) LOW-LEVEL CROSS-SESSION EPISTEMIC INTELLIGENCE/ANALYSIS - End ----

    # -----------------------------------------
    # SAVE STATE AFTER EVERY TURN (PERSISTENCE)
    # -----------------------------------------
    save_state(session_id, state)

    # -----------------------------
    # 🧠 EPISTEMIC (BACKGROUND)
    # -----------------------------
    jung_history_len = len(state.agent_histories.get("jung", []))

    # ✅ Throttle → every 2 turns
    if jung_history_len % 2 == 0:
        background_tasks.add_task(run_epistemic_witness, state, session_id)

    # -----------------------------
    # 📦 ATTACH EPISTEMIC TO RESPONSE
    # -----------------------------
    if hasattr(state, "witness_logs"):
        output["witness_updates"] = state.witness_logs

    print("\n📤 RESPONSE TO UI:", output)

    # -----------------------------
    # CLEANUP
    # -----------------------------
    if output["type"] == "end":
        del sessions[session_id]

    return output

# ---- GET SESSION DATA, PAST CONVERSATION ---- 
@app.get("/session/{session_id}")
def get_session(session_id: str):

    data = load_state(session_id)

    if not data:
        return {"error": "Session not found"}

    return data



@app.get("/health")
def health_check():

    #import requests

    ollama_status = False

    try:
        requests.get("http://127.0.0.1:11434")
        ollama_status = True
    except:
        pass

    return {
        "api": "running",
        "ollama": "connected" if ollama_status else "not running",
        "agents": "ready"
    }