Use Cases

Explore practical use cases for Dynamic Context and learn how it improves character responsiveness, relevance, and contextual awareness in conversations.

Beginner

Personalizing the Conversation

The most common use: telling the character who it is speaking with before any dialogue begins.

Scenario: The player creates a character and enters the world. The NPC should greet them by name and acknowledge their class.

using System.Collections;
using Convai.Runtime.Components;
using Convai.Runtime.Room;
using UnityEngine;

/// <summary>
/// Injects player identity into the active Convai session at session start.
/// The character receives this context before the first exchange and can
/// greet the player by name and class without any dashboard reconfiguration.
/// </summary>
public class PlayerGreeting : MonoBehaviour
{
    [SerializeField] private string _playerName = "Aria";
    [SerializeField] private string _playerClass = "Ranger";

    private IConvaiRoomConnectionService _connectionService;

    private IEnumerator Start()
    {
        while (true)
        {
            ConvaiManager manager = ConvaiManager.ActiveManager;
            if (manager != null && manager.TryGetRoomConnectionService(out _connectionService))
                break;
            yield return null;
        }

        if (_connectionService.IsConnected)
            SendGreeting();
        else
            _connectionService.Connected += OnConnected;
    }

    private void OnConnected()
    {
        _connectionService.Connected -= OnConnected;
        SendGreeting();
    }

    private void SendGreeting()
    {
        string context =
            $"You are speaking with {_playerName}, a level 1 {_playerClass}. " +
            "Greet them warmly and welcome them to the world.";

        // replace — ensures no stale context from a previous session carries over.
        // runLlm: true — triggers an immediate greeting response.
        bool sent = _connectionService.UpdateDynamicContext(context, mode: "replace", runLlm: "true");
        if (!sent)
            Debug.LogWarning("[PlayerGreeting] UpdateDynamicContext returned false.");
    }

    private void OnDestroy()
    {
        if (_connectionService != null)
            _connectionService.Connected -= OnConnected;
    }
}

replace ensures no stale context from a previous session carries over. runLlm: "true" triggers an immediate greeting response.

Injecting the Learner Profile

At session start, push the learner's name, role, and current course. Every character in the session is immediately personalized without any dashboard reconfiguration.

using System.Collections;
using Convai.Runtime.Components;
using Convai.Runtime.Room;
using UnityEngine;

/// <summary>
/// Injects the active learner's profile into the Convai session at session start.
/// Personalizes every character in the session without dashboard reconfiguration.
/// </summary>
public class LearnerProfileInjector : MonoBehaviour
{
    private IConvaiRoomConnectionService _connectionService;

    private IEnumerator Start()
    {
        while (true)
        {
            ConvaiManager manager = ConvaiManager.ActiveManager;
            if (manager != null && manager.TryGetRoomConnectionService(out _connectionService))
                break;
            yield return null;
        }

        if (_connectionService.IsConnected)
            InjectProfile();
        else
            _connectionService.Connected += OnConnected;
    }

    private void OnConnected()
    {
        _connectionService.Connected -= OnConnected;
        InjectProfile();
    }

    private void InjectProfile()
    {
        LearnerProfile profile = LearnerProfileService.Current;
        if (profile == null)
        {
            Debug.LogWarning("[LearnerProfileInjector] No active learner profile found.");
            return;
        }

        string context =
            $"You are speaking with {profile.FullName}, a {profile.JobTitle} " +
            $"in the {profile.Department} department. " +
            $"They are a {profile.ExperienceLevel} learner completing " +
            $"the '{profile.CurrentCourseName}' course. " +
            $"Their stated learning goal: {profile.LearningGoal}. " +
            "Address them by name and calibrate explanations to their experience level.";

        // replace — establishes a clean session baseline.
        // runLlm: false — profile is background context, not a prompt for an immediate response.
        bool sent = _connectionService.UpdateDynamicContext(context, mode: "replace", runLlm: "false");
        if (!sent)
            Debug.LogWarning("[LearnerProfileInjector] UpdateDynamicContext returned false.");
    }

    private void OnDestroy()
    {
        if (_connectionService != null)
            _connectionService.Connected -= OnConnected;
    }
}

Module Progression

As the trainee advances to a new module, replace the context with the module's objectives. The character immediately knows what to focus on — no reconfiguration required.

using Convai.Runtime.Room;
using UnityEngine;

/// <summary>
/// Updates the character's context whenever the trainee advances to a new training module.
/// The character's focus shifts entirely to the new module — no reconfiguration required.
/// </summary>
public class TrainingModuleController : MonoBehaviour
{
    private IConvaiRoomConnectionService _connectionService;

    /// <summary>
    /// Initializes this component with an active room connection service.
    /// Must be called by the session owner after the Convai session is established.
    /// </summary>
    public void Initialize(IConvaiRoomConnectionService connectionService)
    {
        _connectionService = connectionService;
    }

    /// <summary>
    /// Replaces the character's module context with the given module's objectives and concepts.
    /// </summary>
    public void AdvanceToModule(TrainingModule module)
    {
        string context =
            $"The learner is now starting Module {module.Number}: '{module.Title}'. " +
            $"Learning objectives: {string.Join("; ", module.Objectives)}. " +
            $"Key concepts: {string.Join(", ", module.KeyConcepts)}. " +
            $"Estimated duration: {module.EstimatedMinutes} minutes. " +
            "Focus all guidance and questions on this module's content only.";

        // replace — prior module context is no longer relevant.
        bool sent = _connectionService?.UpdateDynamicContext(context, mode: "replace", runLlm: "false") ?? false;
        if (!sent)
            Debug.LogWarning($"[TrainingModuleController] Failed to advance to module '{module.Title}'.");
    }
}

Language Tutor Proficiency Level

A language learning tutor adjusts its vocabulary, sentence complexity, and correction style based on the learner's current proficiency level. A single call at session start — or whenever the level changes.

using Convai.Runtime.Room;
using UnityEngine;

/// <summary>
/// Sets the language tutor's vocabulary and correction style based on the learner's CEFR level.
/// Call at session start and whenever the level changes (e.g. after a level-up assessment).
/// </summary>
public class LanguageTutorLevelManager : MonoBehaviour
{
    private IConvaiRoomConnectionService _connectionService;

    /// <summary>
    /// Initializes this component with an active room connection service.
    /// Must be called by the session owner after the Convai session is established.
    /// </summary>
    public void Initialize(IConvaiRoomConnectionService connectionService)
    {
        _connectionService = connectionService;
    }

    /// <param name="cefrLevel">CEFR proficiency level: A1, A2, B1, B2, C1, or C2.</param>
    /// <param name="nativeLanguage">The learner's native language.</param>
    /// <param name="targetLanguage">The language the learner is studying.</param>
    public void SetProficiencyLevel(string cefrLevel, string nativeLanguage, string targetLanguage)
    {
        string context =
            $"The learner's native language is {nativeLanguage}. " +
            $"They are learning {targetLanguage} at CEFR level {cefrLevel}. " +
            GetLevelInstructions(cefrLevel);

        bool sent = _connectionService?.UpdateDynamicContext(context, mode: "replace", runLlm: "false") ?? false;
        if (!sent)
            Debug.LogWarning($"[LanguageTutorLevelManager] Failed to set proficiency level '{cefrLevel}'.");
    }

    private static string GetLevelInstructions(string level) => level switch
    {
        "A1" or "A2" =>
            "Use only simple present and past tense. Keep sentences under ten words. " +
            "Avoid idioms. Repeat key vocabulary. Correct errors gently by modeling the correct form.",

        "B1" or "B2" =>
            "Use varied sentence structures and introduce common idioms. " +
            "Gently correct grammar errors by rephrasing. Encourage longer responses.",

        "C1" or "C2" =>
            "Speak at natural pace with nuanced vocabulary and idiomatic expressions. " +
            "Correct only significant errors. Engage in abstract or opinion-based discussion.",

        _ => "Adapt to the learner's apparent comprehension level."
    };
}

Injecting Emotional State

Dynamic Context is not limited to facts. Plain-language instructions work equally well for shaping character tone and behavior on the fly.

Scenario: A companion NPC's mood should shift based on what is happening in the scene — calm before a battle, tense during it, relieved after.

using Convai.Runtime.Room;
using UnityEngine;

/// <summary>
/// Controls the companion NPC's emotional state and behavioral tone at runtime.
/// Uses plain-language instructions — no separate emotion pipeline required.
/// The new tone surfaces naturally in the next exchange; no immediate response is triggered.
/// </summary>
public class CompanionMoodController : MonoBehaviour
{
    private IConvaiRoomConnectionService _connectionService;

    /// <summary>
    /// Initializes this component with an active room connection service.
    /// Must be called by the session owner after the Convai session is established.
    /// </summary>
    public void Initialize(IConvaiRoomConnectionService connectionService)
    {
        _connectionService = connectionService;
    }

    /// <summary>
    /// Replaces the character's current emotional state with the given instruction.
    /// </summary>
    /// <param name="moodInstruction">
    /// A plain-language behavioral instruction.
    /// Example: "You sense danger nearby. Speak with urgency and keep responses short."
    /// </param>
    public void SetMood(string moodInstruction)
    {
        // replace — the current mood replaces the previous one; states do not accumulate.
        // runLlm: false — mood is background context; the character should not react unprompted.
        bool sent = _connectionService?.UpdateDynamicContext(
            moodInstruction,
            mode: "replace",
            runLlm: "false") ?? false;

        if (!sent)
            Debug.LogWarning("[CompanionMoodController] UpdateDynamicContext returned false.");
    }
}

// Usage:
// _moodController.SetMood("You sense danger nearby. Speak with urgency and keep responses short.");
// _moodController.SetMood("The threat has passed. You are relieved but still alert.");

runLlm: "false" updates the character's disposition silently — the new tone will surface naturally in the next exchange rather than triggering an unsolicited remark.

Intermediate

Reactive Game State

As the player progresses, the world changes. Dynamic Context lets those changes reach the character in real time.

Scenario: A shopkeeper NPC should always be aware of the player's current gold, recent events, and completed quests.

using Convai.Runtime.Room;
using UnityEngine;

/// <summary>
/// Keeps a shopkeeper NPC's context synchronized with the player's real-time game state.
/// Passive state changes (gold, world events) are sent silently.
/// Significant events (quest completions) trigger an immediate character reaction.
/// </summary>
public class ShopkeeperContextUpdater : MonoBehaviour
{
    private IConvaiRoomConnectionService _connectionService;

    private void Awake()
    {
        PlayerInventory.OnGoldChanged += OnGoldChanged;
        QuestManager.OnQuestCompleted += OnQuestCompleted;
        WorldEvents.OnMajorEventOccurred += OnWorldEvent;
    }

    private void OnDestroy()
    {
        PlayerInventory.OnGoldChanged -= OnGoldChanged;
        QuestManager.OnQuestCompleted -= OnQuestCompleted;
        WorldEvents.OnMajorEventOccurred -= OnWorldEvent;
    }

    /// <summary>
    /// Initializes this component and pushes an initial state snapshot.
    /// Must be called by the session owner after the Convai session is established.
    /// </summary>
    public void Initialize(IConvaiRoomConnectionService connectionService)
    {
        if (connectionService == null)
        {
            Debug.LogError("[ShopkeeperContextUpdater] connectionService must not be null.");
            return;
        }

        _connectionService = connectionService;

        // Push initial snapshot — replace any context from a previous session.
        string snapshot =
            $"Player gold: {PlayerInventory.Gold}g. " +
            $"Completed quests: {string.Join(", ", QuestManager.CompletedQuestNames)}. " +
            $"Merchant Guild reputation: {PlayerStats.MerchantReputation}/100.";

        bool sent = _connectionService.UpdateDynamicContext(snapshot, mode: "replace", runLlm: "false");
        if (!sent)
            Debug.LogWarning("[ShopkeeperContextUpdater] Failed to send initial snapshot.");
    }

    private void OnGoldChanged(int newGold)
    {
        // Silent append — gold amount colors future replies without prompting an unsolicited remark.
        bool sent = _connectionService?.UpdateDynamicContext(
            $"Player gold updated: {newGold}g.",
            mode: "append",
            runLlm: "false") ?? false;

        if (!sent)
            Debug.LogWarning("[ShopkeeperContextUpdater] Failed to update gold context.");
    }

    private void OnQuestCompleted(string questName)
    {
        // runLlm: true — the player returning with a completed quest warrants a reaction.
        bool sent = _connectionService?.UpdateDynamicContext(
            $"The player just completed '{questName}' and is returning to collect their reward.",
            mode: "append",
            runLlm: "true") ?? false;

        if (!sent)
            Debug.LogWarning("[ShopkeeperContextUpdater] Failed to send quest completion context.");
    }

    private void OnWorldEvent(string eventDescription)
    {
        // Silent append — world news enriches future conversation without interrupting the current flow.
        bool sent = _connectionService?.UpdateDynamicContext(
            $"World event: {eventDescription}",
            mode: "append",
            runLlm: "false") ?? false;

        if (!sent)
            Debug.LogWarning("[ShopkeeperContextUpdater] Failed to send world event context.");
    }
}

Passive state (gold, world news) uses runLlm: "false" — it colors future replies without prompting unsolicited comments. Quest completions use runLlm: "true" because a reaction is expected.

Silent Event Accumulation

In high-frequency event scenarios — rapid loot drops, kill streaks, bonus triggers — reacting to every event creates noise. The correct pattern is to accumulate silently and respond only when something meaningful happens, or when the user initiates conversation.

Scenario: A combat commentator tracks kill events throughout a fight. The character stays quiet during the action but has full awareness of everything that happened when the player finally speaks.

using Convai.Runtime.Room;
using UnityEngine;

/// <summary>
/// Accumulates combat events silently in the character's context.
/// The character stays quiet during the action but has full awareness of everything
/// that happened. It speaks only when a meaningful cluster of events is reached
/// or when a combat phase ends naturally.
/// </summary>
public class CombatEventAccumulator : MonoBehaviour
{
    [SerializeField]
    [Tooltip("Number of events to accumulate before triggering a character response.")]
    private int _responseThreshold = 5;

    private IConvaiRoomConnectionService _connectionService;
    private int _pendingEventCount;

    /// <summary>
    /// Initializes this component with an active room connection service.
    /// Must be called by the session owner after the Convai session is established.
    /// </summary>
    public void Initialize(IConvaiRoomConnectionService connectionService)
    {
        if (connectionService == null)
        {
            Debug.LogError("[CombatEventAccumulator] connectionService must not be null.");
            return;
        }
        _connectionService = connectionService;
    }

    /// <summary>
    /// Records a kill event. Accumulates silently until the response threshold is crossed.
    /// </summary>
    public void OnEnemyDefeated(string enemyName, string weaponUsed)
    {
        _pendingEventCount++;

        // Accumulate silently — the character is aware but does not interrupt the action.
        _connectionService?.UpdateDynamicContext(
            $"Player defeated {enemyName} using {weaponUsed}.",
            mode: "append",
            runLlm: "false");

        if (_pendingEventCount < _responseThreshold)
            return;

        // Threshold crossed — a meaningful cluster of events warrants acknowledgment.
        bool sent = _connectionService?.UpdateDynamicContext(
            "The player has been on a significant kill streak. Acknowledge their performance briefly.",
            mode: "append",
            runLlm: "true") ?? false;

        if (!sent)
            Debug.LogWarning("[CombatEventAccumulator] Failed to trigger kill streak response.");

        _pendingEventCount = 0;
    }

    /// <summary>
    /// Flushes any remaining accumulated events at the end of a combat phase.
    /// </summary>
    public void OnCombatPhaseEnded()
    {
        if (_pendingEventCount == 0) return;

        bool sent = _connectionService?.UpdateDynamicContext(
            "Combat phase ended. Summarize what just happened and offer a brief assessment.",
            mode: "append",
            runLlm: "true") ?? false;

        if (!sent)
            Debug.LogWarning("[CombatEventAccumulator] Failed to trigger phase-end response.");

        _pendingEventCount = 0;
    }
}

This keeps the character's intervention meaningful. It has full awareness of the event stream but speaks at deliberate moments rather than reacting to every trigger.

Performance-Driven Coaching Difficulty

A sales training simulator tracks the trainee's real-time performance score. When the score shifts enough to cross a difficulty tier, the character's coaching strategy updates mid-session — no restart needed.

using Convai.Runtime.Room;
using UnityEngine;

/// <summary>
/// Adapts the AI sales trainer's coaching difficulty in real time based on the
/// trainee's performance score. When the score crosses a tier boundary, the
/// character's strategy updates mid-session — no restart required.
/// </summary>
public class SalesTrainerPerformanceAdapter : MonoBehaviour
{
    [SerializeField] private float _easyThreshold = 75f;
    [SerializeField] private float _hardThreshold = 40f;

    private IConvaiRoomConnectionService _connectionService;
    private string _currentDifficultyTier = "normal";

    /// <summary>
    /// Initializes this component with an active room connection service.
    /// Must be called by the session owner after the Convai session is established.
    /// </summary>
    public void Initialize(IConvaiRoomConnectionService connectionService)
    {
        if (connectionService == null)
        {
            Debug.LogError("[SalesTrainerPerformanceAdapter] connectionService must not be null.");
            return;
        }
        _connectionService = connectionService;
    }

    /// <summary>
    /// Evaluates the trainee's current score and updates the coaching strategy
    /// only when the difficulty tier changes or a new skill gap is identified.
    /// </summary>
    /// <param name="newScore">Current performance score (0–100).</param>
    /// <param name="lastMissedSkill">The most recently identified skill gap, or null if none.</param>
    public void OnPerformanceScoreUpdated(float newScore, string lastMissedSkill)
    {
        string newTier = newScore >= _easyThreshold ? "easy"
            : newScore <= _hardThreshold ? "hard"
            : "normal";

        bool tierChanged = newTier != _currentDifficultyTier;
        bool hasNewSkillGap = !string.IsNullOrEmpty(lastMissedSkill);

        // Only push an update when something meaningful has changed.
        if (!tierChanged && !hasNewSkillGap)
            return;

        _currentDifficultyTier = newTier;

        string difficultyInstruction = newTier switch
        {
            "easy" => "The trainee is performing well. Increase objection complexity. Introduce edge cases.",
            "hard" => "The trainee is struggling. Simplify objections. Give conversational hints. Be encouraging.",
            _ => "Maintain current difficulty. Keep objections realistic."
        };

        string context = $"Trainee performance score: {newScore:F0}/100. {difficultyInstruction}";

        if (hasNewSkillGap)
            context += $" The trainee has been missing: {lastMissedSkill}. Subtly probe that area.";

        // replace — the new difficulty state supersedes the previous one entirely.
        bool sent = _connectionService?.UpdateDynamicContext(context, mode: "replace", runLlm: "false") ?? false;
        if (!sent)
            Debug.LogWarning("[SalesTrainerPerformanceAdapter] Failed to update coaching context.");
    }
}

Branching Compliance Scenario State

A compliance training simulation where each trainee decision advances a scenario branch. The AI colleague NPC always has full awareness of what decisions were made, their consequences, and the accumulated compliance score.

using System.Collections.Generic;
using Convai.Runtime.Room;
using UnityEngine;

/// <summary>
/// Tracks trainee decisions in a branching compliance scenario.
/// The AI colleague NPC always has full awareness of the current scenario branch,
/// accumulated decision history, and compliance score — without reconfiguration
/// between decisions.
/// </summary>
public class ComplianceScenarioTracker : MonoBehaviour
{
    private IConvaiRoomConnectionService _connectionService;
    private readonly List<string> _decisionLog = new();
    private string _currentBranch;
    private int _complianceScore = 100;

    /// <summary>
    /// Initializes this component with an active room connection service.
    /// Must be called by the session owner after the Convai session is established.
    /// </summary>
    public void Initialize(IConvaiRoomConnectionService connectionService)
    {
        if (connectionService == null)
        {
            Debug.LogError("[ComplianceScenarioTracker] connectionService must not be null.");
            return;
        }
        _connectionService = connectionService;
    }

    /// <summary>
    /// Records a trainee decision, advances the scenario branch, and pushes the
    /// updated state. Impacts below -20 are treated as critical and trigger an
    /// immediate character reaction.
    /// </summary>
    public void OnTraineeDecision(ScenarioDecision decision)
    {
        _complianceScore += decision.ComplianceImpact;
        _currentBranch = decision.ResultingBranchName;
        _decisionLog.Add($"'{decision.Description}' ({decision.ComplianceImpact:+#;-#;0} pts)");

        bool isError = decision.ComplianceImpact < 0;
        bool isCriticalError = decision.ComplianceImpact < -20;

        string context =
            $"Current scenario branch: {_currentBranch}. " +
            $"Trainee compliance score: {_complianceScore}/100. " +
            $"Decision history: {string.Join(" → ", _decisionLog)}. " +
            $"Last decision consequence: {decision.ConsequenceDescription}. " +
            (isError
                ? "The trainee made a compliance error. Reference it naturally — do not lecture."
                : "The trainee made the correct compliance decision. Acknowledge their judgment.");

        // replace — the full scenario state replaces the previous snapshot on every decision.
        bool sent = _connectionService?.UpdateDynamicContext(
            context,
            mode: "replace",
            runLlm: isCriticalError ? "true" : "false") ?? false;

        if (!sent)
            Debug.LogWarning("[ComplianceScenarioTracker] Failed to update scenario context.");
    }

    /// <summary>
    /// Signals scenario completion and triggers a structured debrief from the character.
    /// </summary>
    public void OnScenarioCompleted()
    {
        string debrief =
            $"The scenario is complete. Final compliance score: {_complianceScore}/100. " +
            $"Provide a brief debrief: acknowledge strengths, identify the most significant error, " +
            $"and give one actionable recommendation. " +
            $"Decision log: {string.Join("; ", _decisionLog)}.";

        bool sent = _connectionService?.UpdateDynamicContext(debrief, mode: "append", runLlm: "true") ?? false;
        if (!sent)
            Debug.LogWarning("[ComplianceScenarioTracker] Failed to send scenario debrief.");
    }
}

Medical Simulation Patient Vitals Monitor

A clinical training simulation where a patient NPC's vitals change in real time. The character absorbs routine fluctuations silently but escalates when values cross clinically significant thresholds — without spoiling the diagnosis for the trainee.

using System.Collections.Generic;
using Convai.Runtime.Room;
using UnityEngine;

/// <summary>
/// Synchronizes real-time patient vitals with the Convai character context.
/// The character absorbs routine fluctuations silently and escalates only when
/// values cross clinically significant thresholds — without spoiling the diagnosis.
/// </summary>
public class PatientVitalsSimulator : MonoBehaviour
{
    // Clinical thresholds for escalation triggers.
    private const float HypotensionThreshold = 90f;
    private const int BradycardiaThreshold = 50;
    private const float HypoxiaThreshold = 90f;

    private IConvaiRoomConnectionService _connectionService;
    private bool _criticalStateActive;

    /// <summary>
    /// Initializes this component with an active room connection service.
    /// Must be called by the session owner after the Convai session is established.
    /// </summary>
    public void Initialize(IConvaiRoomConnectionService connectionService)
    {
        if (connectionService == null)
        {
            Debug.LogError("[PatientVitalsSimulator] connectionService must not be null.");
            return;
        }
        _connectionService = connectionService;
    }

    /// <summary>
    /// Pushes updated patient vitals to the character context.
    /// Triggers an escalation or recovery response when clinical thresholds are crossed.
    /// </summary>
    public void OnVitalsUpdated(PatientVitals vitals)
    {
        // Always update silently — replace, not append (vitals are current state, not a log).
        string vitalsContext =
            $"Current patient vitals — " +
            $"BP: {vitals.SystolicBP}/{vitals.DiastolicBP} mmHg, " +
            $"HR: {vitals.HeartRate} bpm, " +
            $"SpO2: {vitals.OxygenSaturation}%, " +
            $"Temp: {vitals.Temperature:F1}°C, " +
            $"RR: {vitals.RespiratoryRate}/min.";

        _connectionService?.UpdateDynamicContext(vitalsContext, mode: "replace", runLlm: "false");

        bool isCritical =
            vitals.SystolicBP < HypotensionThreshold ||
            vitals.HeartRate < BradycardiaThreshold ||
            vitals.OxygenSaturation < HypoxiaThreshold;

        if (isCritical && !_criticalStateActive)
        {
            _criticalStateActive = true;

            bool sent = _connectionService?.UpdateDynamicContext(
                $"CLINICAL ALERT: {BuildCriticalFindings(vitals)}. " +
                "React as a deteriorating patient. Express distress appropriate to severity. " +
                "Do not tell the trainee what to do — let them diagnose and act.",
                mode: "append",
                runLlm: "true") ?? false;

            if (!sent)
                Debug.LogWarning("[PatientVitalsSimulator] Failed to send critical escalation.");
        }
        else if (!isCritical && _criticalStateActive)
        {
            _criticalStateActive = false;

            bool sent = _connectionService?.UpdateDynamicContext(
                "Patient vitals have stabilized. Express relief. Let the trainee assess next steps.",
                mode: "append",
                runLlm: "true") ?? false;

            if (!sent)
                Debug.LogWarning("[PatientVitalsSimulator] Failed to send stabilization context.");
        }
    }

    private static string BuildCriticalFindings(PatientVitals vitals)
    {
        var findings = new List<string>();

        if (vitals.SystolicBP < HypotensionThreshold)
            findings.Add($"hypotension (BP {vitals.SystolicBP}/{vitals.DiastolicBP})");
        if (vitals.HeartRate < BradycardiaThreshold)
            findings.Add($"bradycardia (HR {vitals.HeartRate} bpm)");
        if (vitals.OxygenSaturation < HypoxiaThreshold)
            findings.Add($"hypoxia (SpO2 {vitals.OxygenSaturation}%)");

        return string.Join(" and ", findings);
    }
}

Advanced

Cross-Scenario Competency Gap Tracker

A multi-scenario training platform tracks which competencies the learner has demonstrated or missed across all scenarios in a session. Each new scenario is informed by the accumulated competency profile so the AI can naturally probe weak areas without the learner being aware of it.

using System.Collections.Generic;
using System.Linq;
using Convai.Runtime.Room;
using UnityEngine;

/// <summary>
/// Tracks learner competency observations across multiple training scenarios.
/// Each new scenario is informed by the accumulated competency profile, allowing
/// the AI to naturally probe weak areas without the learner being aware of it.
/// Persists across scene loads via DontDestroyOnLoad.
/// </summary>
public sealed class CompetencyGapTracker : MonoBehaviour
{
    public static CompetencyGapTracker Instance { get; private set; }

    private IConvaiRoomConnectionService _connectionService;
    private readonly Dictionary<string, CompetencyRecord> _competencies = new();

    private void Awake()
    {
        if (Instance != null && Instance != this)
        {
            Destroy(gameObject);
            return;
        }
        Instance = this;
        DontDestroyOnLoad(gameObject);
    }

    /// <summary>
    /// Initializes this tracker with an active room connection service.
    /// Must be called by the session owner after the Convai session is established.
    /// </summary>
    public void Initialize(IConvaiRoomConnectionService connectionService)
    {
        if (connectionService == null)
        {
            Debug.LogError("[CompetencyGapTracker] connectionService must not be null.");
            return;
        }
        _connectionService = connectionService;
    }

    /// <summary>
    /// Records a pass or fail observation for a named competency.
    /// Requires at least two observations before a competency influences scenario injection.
    /// </summary>
    /// <param name="competencyId">Stable identifier (e.g. "active-listening").</param>
    /// <param name="competencyName">Human-readable name passed to the character.</param>
    /// <param name="demonstrated">True if the trainee demonstrated this competency.</param>
    public void RecordObservation(string competencyId, string competencyName, bool demonstrated)
    {
        if (!_competencies.TryGetValue(competencyId, out var record))
            record = _competencies[competencyId] = new CompetencyRecord(competencyName);

        record.AddObservation(demonstrated);
    }

    /// <summary>
    /// Pushes accumulated competency intelligence into the next scenario's context.
    /// The character will naturally create opportunities for the trainee to practice gap areas.
    /// </summary>
    public void OnNewScenarioStarted(string scenarioName, string scenarioContext)
    {
        string gaps = BuildGapSummary();
        string strengths = BuildStrengthSummary();

        string context =
            $"New scenario: '{scenarioName}'. {scenarioContext} " +
            $"Trainee strengths from prior scenarios: {strengths}. " +
            $"Competency gaps to probe this scenario: {gaps}. " +
            $"Create natural opportunities to practice gap areas. " +
            "Do not make it obvious you are targeting weaknesses.";

        // replace — each scenario starts with a fresh, scenario-specific context.
        bool sent = _connectionService?.UpdateDynamicContext(context, mode: "replace", runLlm: "false") ?? false;
        if (!sent)
            Debug.LogWarning($"[CompetencyGapTracker] Failed to inject context for scenario '{scenarioName}'.");
    }

    /// <summary>
    /// Triggers a full session debrief based on the complete competency record.
    /// Call when all scenarios in the session are complete.
    /// </summary>
    public void OnSessionCompleted()
    {
        string report = string.Join("; ", _competencies.Values
            .OrderBy(r => r.SuccessRate)
            .Select(r => $"{r.Name}: {r.SuccessRate:P0} ({r.TotalObservations} observations)"));

        bool sent = _connectionService?.UpdateDynamicContext(
            "Training session complete. Deliver a structured debrief: overall performance summary, " +
            "two key strengths, two priority development areas, one concrete next step. " +
            $"Competency data: {report}",
            mode: "replace",
            runLlm: "true") ?? false;

        if (!sent)
            Debug.LogWarning("[CompetencyGapTracker] Failed to trigger session debrief.");
    }

    // Competencies with fewer than two observations are excluded — insufficient data.
    private string BuildGapSummary()
    {
        var gaps = _competencies.Values
            .Where(r => r.TotalObservations >= 2 && r.SuccessRate < 0.6f)
            .Select(r => $"{r.Name} ({r.SuccessRate:P0})")
            .ToList();

        return gaps.Count > 0 ? string.Join(", ", gaps) : "No significant gaps identified yet.";
    }

    private string BuildStrengthSummary()
    {
        var strengths = _competencies.Values
            .Where(r => r.TotalObservations >= 2 && r.SuccessRate >= 0.8f)
            .Select(r => $"{r.Name} ({r.SuccessRate:P0})")
            .ToList();

        return strengths.Count > 0 ? string.Join(", ", strengths) : "Insufficient data.";
    }

    private sealed class CompetencyRecord
    {
        public string Name { get; }
        public int TotalObservations { get; private set; }
        private int _demonstratedCount;

        public float SuccessRate => TotalObservations == 0 ? 0f
            : (float)_demonstratedCount / TotalObservations;

        public CompetencyRecord(string name) => Name = name;

        public void AddObservation(bool demonstrated)
        {
            TotalObservations++;
            if (demonstrated) _demonstratedCount++;
        }
    }
}

Usage across scenarios:

// During a scenario — record what you observe
CompetencyGapTracker.Instance.RecordObservation(
    "active-listening", "Active Listening", demonstrated: true);

CompetencyGapTracker.Instance.RecordObservation(
    "objection-handling", "Objection Handling", demonstrated: false);

// Inject accumulated intelligence before the next scenario
CompetencyGapTracker.Instance.OnNewScenarioStarted(
    "Difficult Customer — Refund Request",
    "The customer is upset about a delayed order and requesting a full refund.");

// At session end — trigger full debrief
CompetencyGapTracker.Instance.OnSessionCompleted();

Real-Time Procedural Assessment

A high-fidelity procedural trainer (flight simulator, surgical trainer, industrial safety) monitors every action against a defined checklist. Actions accumulate silently. When the same step fails repeatedly, the AI instructor intervenes with a Socratic question — never giving the answer, but guiding the trainee's thinking.

using System.Collections.Generic;
using System.Linq;
using Convai.Runtime.Room;
using UnityEngine;

/// <summary>
/// AI instructor for high-fidelity procedural training simulations
/// (flight, surgical, industrial safety). Monitors every trainee action silently
/// and intervenes with Socratic guidance only when a repeated error pattern is
/// detected — preserving independent problem-solving throughout.
/// </summary>
public class ProceduralAssessmentInstructor : MonoBehaviour
{
    [SerializeField]
    [Tooltip("Number of failures on the same step before a coaching intervention is triggered.")]
    private int _errorPatternThreshold = 2;

    [SerializeField]
    [Tooltip("Minimum seconds between coaching interventions. Prevents rapid-fire interruptions.")]
    private float _interventionCooldownSeconds = 30f;

    private IConvaiRoomConnectionService _connectionService;
    private readonly Dictionary<string, int> _stepErrorCounts = new();
    private readonly List<string> _completedSteps = new();
    private float _lastInterventionTime = float.MinValue;
    private string _currentProcedureName;

    /// <summary>
    /// Initializes this component with an active room connection service.
    /// Must be called by the session owner after the Convai session is established.
    /// </summary>
    public void Initialize(IConvaiRoomConnectionService connectionService)
    {
        if (connectionService == null)
        {
            Debug.LogError("[ProceduralAssessmentInstructor] connectionService must not be null.");
            return;
        }
        _connectionService = connectionService;
    }

    /// <summary>
    /// Signals the start of a procedure. Resets all tracking state and instructs
    /// the character to observe silently until intervention is warranted.
    /// </summary>
    public void BeginProcedure(string procedureName)
    {
        _currentProcedureName = procedureName;
        _stepErrorCounts.Clear();
        _completedSteps.Clear();
        _lastInterventionTime = float.MinValue;

        bool sent = _connectionService?.UpdateDynamicContext(
            $"The trainee is beginning the '{procedureName}' procedure. " +
            "Observe silently unless asked for help or a critical error occurs. " +
            "Do not volunteer the next step — the trainee must demonstrate independent recall.",
            mode: "replace",
            runLlm: "false") ?? false;

        if (!sent)
            Debug.LogWarning($"[ProceduralAssessmentInstructor] Failed to begin procedure '{procedureName}'.");
    }

    /// <summary>
    /// Records a step attempt. Correct steps are acknowledged silently.
    /// Repeated failures on the same step trigger a Socratic intervention
    /// once the error threshold is met and the cooldown has elapsed.
    /// </summary>
    /// <param name="stepId">Stable identifier for this step.</param>
    /// <param name="stepDescription">Human-readable description passed to the character.</param>
    /// <param name="wasCorrect">Whether the step was performed correctly.</param>
    /// <param name="wasInSequence">Whether the step was performed in the correct order.</param>
    public void OnStepAttempted(string stepId, string stepDescription, bool wasCorrect, bool wasInSequence)
    {
        if (wasCorrect && wasInSequence)
        {
            _completedSteps.Add(stepDescription);

            // Silently acknowledge correct progress — no character response needed.
            _connectionService?.UpdateDynamicContext(
                $"Step completed correctly: {stepDescription}. " +
                $"Progress: {_completedSteps.Count} steps done.",
                mode: "append",
                runLlm: "false");

            return;
        }

        _stepErrorCounts.TryGetValue(stepId, out int errorCount);
        _stepErrorCounts[stepId] = ++errorCount;

        // Accumulate the error silently for context grounding.
        _connectionService?.UpdateDynamicContext(
            $"Error on step '{stepDescription}' " +
            $"(error count: {errorCount}, in-sequence: {wasInSequence}).",
            mode: "append",
            runLlm: "false");

        bool patternDetected = errorCount >= _errorPatternThreshold;
        bool cooldownElapsed = Time.realtimeSinceStartup - _lastInterventionTime > _interventionCooldownSeconds;

        if (!patternDetected || !cooldownElapsed)
            return;

        _lastInterventionTime = Time.realtimeSinceStartup;

        // Intervene with a Socratic question — guide thinking without revealing the answer.
        bool sent = _connectionService?.UpdateDynamicContext(
            $"The trainee has failed '{stepDescription}' {errorCount} times. " +
            "Intervene with one Socratic question that guides their thinking without revealing the answer. " +
            "Be calm and constructive.",
            mode: "append",
            runLlm: "true") ?? false;

        if (!sent)
            Debug.LogWarning("[ProceduralAssessmentInstructor] Failed to trigger coaching intervention.");
    }

    /// <summary>
    /// Signals the end of the procedure and triggers a structured debrief from the character.
    /// </summary>
    public void OnProcedureCompleted(bool passed, float elapsedSeconds)
    {
        int totalErrors = _stepErrorCounts.Values.Sum();
        string worstStep = _stepErrorCounts.Count > 0
            ? _stepErrorCounts.OrderByDescending(kv => kv.Value).First().Key
            : "none";

        // replace — the debrief context supersedes all accumulated step-level context.
        bool sent = _connectionService?.UpdateDynamicContext(
            $"Procedure '{_currentProcedureName}' complete. Result: {(passed ? "PASS" : "FAIL")}. " +
            $"Time: {elapsedSeconds:F0}s. Total errors: {totalErrors}. " +
            $"Most problematic step: '{worstStep}' ({_stepErrorCounts.GetValueOrDefault(worstStep, 0)} errors). " +
            "Deliver a one-sentence overall assessment, the single most important error to address, and one thing they did well.",
            mode: "replace",
            runLlm: "true") ?? false;

        if (!sent)
            Debug.LogWarning("[ProceduralAssessmentInstructor] Failed to trigger procedure debrief.");
    }
}

What this pattern enforces:

The instructor has full awareness of every action without constantly interrupting — the character is always listening
The Socratic threshold prevents the instructor from becoming a hint machine — it speaks only when a genuine learning barrier is detected
The cooldown prevents rapid-fire interventions that would undermine independent thinking
The final debrief draws on the accumulated event log for a substantive, evidence-based assessment rather than a generic summary

Reactive Character with Autonomous Response Thresholds

The most sophisticated pattern: the character monitors a continuous stream of game state updates and decides autonomously when to speak — without waiting for the user to initiate. This is a reactive architecture. The character is always listening, but only surfaces when something crosses a meaningful threshold.

Scenario: A companion NPC tracks the player's health passively throughout the session, warns at a moderate drop, and urgently intervenes at a critical level — all without any input from the player.

using Convai.Runtime.Room;
using UnityEngine;

/// <summary>
/// A companion NPC that monitors game state passively and reacts autonomously
/// when meaningful thresholds are crossed — without waiting for the player to speak.
/// Silent updates keep the character continuously informed; runLlm: true is reserved
/// exclusively for moments that warrant a proactive response.
/// </summary>
public class ReactiveCompanion : MonoBehaviour
{
    [SerializeField]
    [Tooltip("Health percentage at which the companion first expresses concern.")]
    private float _warningThreshold = 50f;

    [SerializeField]
    [Tooltip("Health percentage at which the companion intervenes urgently.")]
    private float _criticalThreshold = 20f;

    private IConvaiRoomConnectionService _connectionService;
    private float _lastHealth = 100f;

    // Flags prevent repeated interventions for the same threshold crossing.
    private bool _warningIssued;
    private bool _criticalIssued;

    /// <summary>
    /// Initializes this component with an active room connection service.
    /// Must be called by the session owner after the Convai session is established.
    /// </summary>
    public void Initialize(IConvaiRoomConnectionService connectionService)
    {
        if (connectionService == null)
        {
            Debug.LogError("[ReactiveCompanion] connectionService must not be null.");
            return;
        }
        _connectionService = connectionService;
    }

    /// <summary>
    /// Silently tracks player health and triggers proactive responses at defined thresholds.
    /// </summary>
    public void OnPlayerHealthChanged(float newHealth)
    {
        // Keep the character silently informed — replace, not append (current state, not a log).
        _connectionService?.UpdateDynamicContext(
            $"Player health: {newHealth:F0}%.",
            mode: "replace",
            runLlm: "false");

        // Warn once when crossing the warning threshold downward.
        if (newHealth <= _warningThreshold && _lastHealth > _warningThreshold && !_warningIssued)
        {
            _warningIssued = true;

            bool sent = _connectionService?.UpdateDynamicContext(
                "Player health has dropped below 50%. Express concern and suggest they find cover or heal.",
                mode: "append",
                runLlm: "true") ?? false;

            if (!sent)
                Debug.LogWarning("[ReactiveCompanion] Failed to send health warning.");
        }

        // Critical intervention — triggers once per threshold crossing until health recovers.
        if (newHealth <= _criticalThreshold && !_criticalIssued)
        {
            _criticalIssued = true;

            bool sent = _connectionService?.UpdateDynamicContext(
                "CRITICAL: Player health is dangerously low. React with urgency — tell them to heal immediately.",
                mode: "append",
                runLlm: "true") ?? false;

            if (!sent)
                Debug.LogWarning("[ReactiveCompanion] Failed to send critical health warning.");
        }

        // Reset both flags when health recovers above the warning threshold.
        if (newHealth > _warningThreshold)
        {
            _warningIssued = false;
            _criticalIssued = false;
        }

        _lastHealth = newHealth;
    }

    /// <summary>
    /// Triggers an immediate proactive warning when the player enters a known danger zone.
    /// </summary>
    public void OnPlayerEnteredDangerZone(string zoneName)
    {
        bool sent = _connectionService?.UpdateDynamicContext(
            $"The player has entered {zoneName}, a known danger zone. Proactively warn them.",
            mode: "append",
            runLlm: "true") ?? false;

        if (!sent)
            Debug.LogWarning("[ReactiveCompanion] Failed to send danger zone warning.");
    }

    /// <summary>
    /// Nudges the character to check in naturally after the player has been idle too long.
    /// Only the intent is injected — the LLM determines the specific wording.
    /// </summary>
    public void OnPlayerIdleTooLong()
    {
        bool sent = _connectionService?.UpdateDynamicContext(
            "The player has been idle for an unusually long time. Check in with them naturally.",
            mode: "append",
            runLlm: "true") ?? false;

        if (!sent)
            Debug.LogWarning("[ReactiveCompanion] Failed to send idle check-in.");
    }
}

Silent updates via runLlm: "false" keep the character's knowledge current at all times. runLlm: "true" is reserved exclusively for moments that genuinely warrant a proactive response. The threshold logic ensures the character intervenes once — not repeatedly — until the situation resolves.

Last updated 25 days ago

Was this helpful?

hashtagBeginner

hashtagPersonalizing the Conversation

hashtagInjecting the Learner Profile

hashtagModule Progression

hashtagLanguage Tutor Proficiency Level

hashtagInjecting Emotional State

hashtagIntermediate

hashtagReactive Game State

hashtagSilent Event Accumulation

hashtagPerformance-Driven Coaching Difficulty

hashtagBranching Compliance Scenario State

hashtagMedical Simulation Patient Vitals Monitor

hashtagAdvanced

hashtagCross-Scenario Competency Gap Tracker

hashtagReal-Time Procedural Assessment

hashtagReactive Character with Autonomous Response Thresholds