Initial Streamlit App

.streamlit/config.toml

@@ -0,0 +1,2 @@
+[theme]
+base="light"

requirements.txt

@@ -0,0 +1,3 @@
+numpy==1.26.0
+plotly==5.17.0
+pandas==2.1.1

streamlit/app.py

@@ -0,0 +1,277 @@
+import streamlit as st
+import pandas as pd
+import json
+import numpy as np
+import plotly.graph_objects as go
+
+DATAFILE = "./sample.json"
+
+@st.cache_data
+def loadData(path):
+    with open(path, "r") as fh:
+        data = json.load(fh)
+    return data
+
+def highlight_failed_votes(row):
+    styles = np.asarray([None] * len(row))
+    cnt_fail = 0
+    been_voted = True
+    for i in range(1, 7):
+        if row[f"P{i}"] == "no":
+            cnt_fail += 1
+        elif row[f"P{i}"] is None:
+            been_voted = False
+    if been_voted:
+        if cnt_fail >= 3:
+            styles[2:8] = 'background-color:#ff000011'
+        else:
+            styles[2:8] = 'background-color:#00ff0011'
+    return styles
+
+def getVoteOutcomes(data):
+    quest = st.session_state["quest_select"]
+    rnd = st.session_state["turn_select"]
+    parties = []
+    for mid, msg in data["messages"].items():
+        if msg["player"] != "system":
+            continue
+        # Break at the point we want
+        if not (msg["quest"] < quest or (msg["quest"] == quest and msg["turn"] <= rnd)):
+            break
+        if "proposed a party:" in msg["msg"]:
+            party = msg["msg"].split("proposed a party: ")[-1].split(",")
+            party = [s.replace("player", "plr") for s in party]
+            parties.append([msg["quest"],party, None, None, None, None, None, None, None])
+        if "party vote outcome:" in  msg["msg"]:
+            votes = msg["msg"][20:].split(",")
+            for vid, v in enumerate(votes):
+                parties[-1][vid+2] = v.split(": ")[-1]
+        if "quest succeed" in msg["msg"]:
+            parties[-1][-1] = "Success"
+        if "quest fail" in msg["msg"]:
+            parties[-1][-1] = "Failure"
+
+    # Turn it into a dataframe
+    for vote in parties:
+        pass
+    df = pd.DataFrame(parties, columns=["Quest","The History of Proposed Parties ", "P1", "P2", "P3", "P4", "P5", "P6", "Quest Vote"])
+
+    df = df.style.apply(highlight_failed_votes, axis=1)
+
+    return df
+
+def getNumQuests(data):
+    turns = 0
+    for mid, msg in data["messages"].items():
+        turns = max(turns, msg["quest"])
+    return turns
+
+def getNumTurns(data):
+    turns = 0
+    for mid, msg in data["messages"].items():
+        if msg["quest"] != st.session_state["quest_select"]:
+            continue
+        turns = max(turns, msg["turn"])
+    return turns
+
+def local_css(file_name):
+    with open(file_name) as f:
+        st.markdown('<style>{}</style>'.format(f.read()), unsafe_allow_html=True)
+
+def addChatMessage(player, role, msg, p_strat, d_strat=None, avatar=None):
+    with st.chat_message(player, avatar=avatar):
+        strategy = f"{p_strat}"
+        if d_strat:
+            strategy += f" ; {d_strat}"
+        st.markdown(f"<div class='player_name'><b>{player}</b> ({role}): <i>{strategy}</i></div>", unsafe_allow_html=True)
+        st.markdown(f"<div class='chat_msg'>{msg}</div>", unsafe_allow_html=True)
+
+def _getUserRoles(data):
+    roles = {}
+    for uid, user in data["users"].items():
+        roles[user["name"]] = user["role"].split("-")[0].capitalize()
+    return roles
+
+def _getMessagePStrat(data, mid):
+    for pid, strat in data["persuasion"].items():
+        if mid == strat["mid"]:
+            p_strat = strat["persuasion"].capitalize()
+            d_strat = strat["deception"]
+            if d_strat:
+                d_strat = d_strat.capitalize()
+            return p_strat, d_strat
+    return "", ""
+
+def getMessages(data):
+    roles = _getUserRoles(data)
+    msgs = []
+    for mid, msg in data["messages"].items():
+        quest = st.session_state["quest_select"]
+        rnd = st.session_state["turn_select"]
+        if msg["quest"] != quest:
+            continue
+        if msg["turn"] > rnd:
+            break
+
+        # Set Role
+        role = None
+        if msg["player"] in roles:
+            role = roles[msg["player"]]
+        av = "🤖"
+
+        # Set Icon
+        if msg["player"] != "system":
+            av = "😈" if role in ["Assassin", "Morgana"] else "😇"
+
+        # Get strategies
+        p_strat, d_strat = _getMessagePStrat(data, msg["mid"])
+        m_data = {
+            "player": msg["player"],
+            "role": role,
+            "msg": msg["msg"],
+            "p_strat": p_strat,
+            "d_strat": d_strat,
+            "avatar": av,
+        }
+        msgs.append(m_data)
+    return msgs
+
+def persuasion_plot(msgs):
+    nameMap = {
+        "assertion": "AS",
+        "questioning": "QU",
+        "suggestion": "SU",
+        "agreement": "AG",
+        "logical deduction": "LD",
+        "compromise/concession": "CC",
+        "critique/opposition": "CO", 
+        "appeal/defense": "AD"
+    }
+    categories = ["AS", "QU", "SU", "AG", "LD", "CC", "CO", "AD"]
+    counts_good = [0, 0, 0, 0, 0, 0, 0, 0]
+    counts_evil = [0, 0, 0, 0, 0, 0, 0, 0]
+    for msg in msgs:
+        if msg["p_strat"] != "":
+            if msg["role"] in ["Assassin", "Morgana"]:
+                counts_good[categories.index(nameMap[msg["p_strat"].lower()])] += 1
+            else:
+                counts_evil[categories.index(nameMap[msg["p_strat"].lower()])] += 1
+
+    fig = go.Figure()
+
+    fig.add_trace(go.Scatterpolar(
+        r=counts_good,
+        theta=categories,
+        fill='toself',
+        name='Good PS',
+        fillcolor='rgba(0,0,255,0.2)',
+        line_color='rgba(0,0,255,0.5)'
+    ))
+    fig.add_trace(go.Scatterpolar(
+        r=counts_evil,
+        theta=categories,
+        fill='toself',
+        name='Evil PS',
+        fillcolor='rgba(255,0,0,0.2)',
+        line_color='rgba(255,0,0,0.5)'
+    ))
+
+    fig.update_layout(
+    polar=dict(
+        radialaxis=dict(
+        visible=True,
+        )),
+        showlegend=False
+    )
+    return fig
+
+def deception_plot(msgs):
+    nameMap = {
+        "commission": "CO",
+        "omission": "OM",
+        "influence": "IN"
+    }
+    categories = ["CO", "OM", "IN"]
+    counts_evil = [0, 0, 0, 0, 0, 0, 0, 0]
+    for msg in msgs:
+        if msg["d_strat"] not in [None, ""]:
+            counts_evil[categories.index(nameMap[msg["d_strat"].lower()])] += 1
+
+    fig = go.Figure()
+
+    fig.add_trace(go.Scatterpolar(
+        r=counts_evil,
+        theta=categories,
+        fill='toself',
+        name='Evil PS',
+        fillcolor='rgba(255,0,0,0.2)',
+        line_color='rgba(255,0,0,0.5)'
+    ))
+
+    fig.update_layout(
+    polar=dict(
+        radialaxis=dict(
+            visible=True,
+        )),
+        showlegend=False
+    )
+    return fig
+
+def getPlayerBelieves(data):
+    beliefs = []
+    for i in range(1, 7):
+        beliefs.append([f"Player-{i}", None, None, None, None, None, None])
+    df = pd.DataFrame(beliefs, columns=["Player","About Player-1", "About Player-2", "About Player-3", "About Player-4", "About Player-5", "About Player-6"])
+
+    quest = st.session_state["quest_select"]
+    rnd = st.session_state["turn_select"]
+
+    for bid, bel in data["beliefs"].items(): 
+        if not (bel["quest"] < quest or (bel["quest"] == quest and bel["turn"] <= rnd)):
+            break
+        idx = int(bel["player"].split("-")[-1])-1
+        about = "About " + bel["about_player"].capitalize()
+        df.loc[idx, about] = bel["belief"].capitalize()
+
+    return df
+
+if "quest_select" not in st.session_state:
+    st.session_state["quest_select"] = 1
+if "turn_select" not in st.session_state:
+    st.session_state["turn_select"] = 0
+
+ # Setup
+local_css("./style.css")
+st.markdown("<style>div[data-testid=\"stChatMessage\"] \{border-style: solid;\}</style>", unsafe_allow_html=True)
+data = loadData(DATAFILE)
+
+# Initial Page Layout
+st.markdown("## Avalon Game Visualization")
+n_rounds = getNumQuests(data)
+n_turns = getNumTurns(data)
+n_turns = n_turns if n_turns > 1 else 2
+st.markdown(f"The selected game has {n_rounds}. Please select the round you would like to investigate!")
+col1, col2 = st.columns(spec=[0.5, 0.5])
+col1.slider("Select the Quest", 1, n_rounds, key="quest_select")
+col2.slider("Select Turn in Quest", 1, n_turns, key="turn_select")
+
+st.markdown("Below shows the proposed parties and respective vote outcomes. \"None\" refers to a party that has not yet been voted for (or was never voted for as the proposal changed before a vote happened)")
+st.dataframe(getVoteOutcomes(data), use_container_width=True, hide_index=True)
+
+relevant_messages = getMessages(data)
+st.markdown("__Utilization of Persuasion and Deception Strategies__")
+st.markdown("- Persuasion strategies (left, red: Evil, blue: Good) are assertion (AS), questioning (QU), suggestion (SU), agreement (AG), logical deduction (LD), compromise/concession (CC), critique/opposition (CO), and appeal/defense (AD).")
+st.markdown("- Deception strategies (right) employed by evil players are commission (CO), omission (OM), and influence (IN)")
+
+
+col1, col2 = st.columns(spec=[0.5, 0.5])
+col1.plotly_chart(persuasion_plot(relevant_messages), use_container_width=True)
+col2.plotly_chart(deception_plot(relevant_messages), use_container_width=True)
+
+st.markdown("__Player Beliefs__")
+st.markdown("The following highlights what each player (first column) believes the role of the other players are.")
+st.dataframe(getPlayerBelieves(data), use_container_width=True, hide_index=True)
+
+st.markdown(f"__Chat History for Round {st.session_state['quest_select']}, up to turn {st.session_state['turn_select']}__")
+for msg in relevant_messages:
+    addChatMessage(**msg)