crates/mem-cli/src/ingest_worker.rs

@@ -34,6 +34,36 @@ impl std::fmt::Display for JobStatus {
     }
 }
 
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    /// Mock JobStatusStore for testing
+    pub struct MockJobStatusStore {
+        updates: std::sync::Arc<std::sync::Mutex<Vec<(String, JobStatus)>>>,
+    }
+
+    impl MockJobStatusStore {
+        pub fn new() -> Self {
+            Self {
+                updates: std::sync::Arc::new(std::sync::Mutex::new(Vec::new())),
+            }
+        }
+
+        pub fn updates(&self) -> Vec<(String, JobStatus)> {
+            self.updates.lock().unwrap().clone()
+        }
+    }
+
+    #[async_trait::async_trait]
+    impl JobStatusStore for MockJobStatusStore {
+        async fn update_status(&self, ingest_id: &str, status: JobStatus) -> Result<()> {
+            self.updates.lock().unwrap().push((ingest_id.to_string(), status));
+            Ok(())
+        }
+    }
+}
+
 /// Structured logging context for ingest operations — ensures consistent field names across all logs
 #[derive(Debug, Clone)]
 pub struct IngestLogContext {
@@ -54,6 +84,36 @@ impl IngestLogContext {
     }
 }
 
+/// Job status store trait — abstracts database persistence of job status (enables mocking)
+#[async_trait::async_trait]
+pub trait JobStatusStore: Send + Sync {
+    /// Update job status in storage
+    async fn update_status(&self, ingest_id: &str, status: JobStatus) -> Result<()>;
+}
+
+/// PostgreSQL implementation of JobStatusStore
+pub struct PgJobStatusStore {
+    pool: PgPool,
+}
+
+impl PgJobStatusStore {
+    pub fn new(pool: PgPool) -> Self {
+        Self { pool }
+    }
+}
+
+#[async_trait::async_trait]
+impl JobStatusStore for PgJobStatusStore {
+    async fn update_status(&self, ingest_id: &str, status: JobStatus) -> Result<()> {
+        sqlx::query("UPDATE ingest_jobs SET status=$1, started_at=NOW() WHERE ingest_id=$2")
+            .bind(status.as_str())
+            .bind(ingest_id)
+            .execute(&self.pool)
+            .await?;
+        Ok(())
+    }
+}
+
 
 /// Ingest worker — processes queued records through entity/fact extraction pipeline
 pub struct IngestWorker {
@@ -61,6 +121,7 @@ pub struct IngestWorker {
     vector_store: Arc<VectorStore>,
     embeddings: Arc<EmbeddingsClient>,
     pipeline: Arc<IngestPipeline>,
+    job_status_store: Arc<dyn JobStatusStore>,
 }
 
 impl IngestWorker {
@@ -68,6 +129,16 @@ impl IngestWorker {
     pub fn new(
         pool: PgPool,
         embeddings: EmbeddingsClient,
+    ) -> Self {
+        let job_status_store = Arc::new(PgJobStatusStore::new(pool.clone()));
+        Self::with_job_store(pool, embeddings, job_status_store)
+    }
+
+    /// Create worker with custom job status store (for testing)
+    pub fn with_job_store(
+        pool: PgPool,
+        embeddings: EmbeddingsClient,
+        job_status_store: Arc<dyn JobStatusStore>,
     ) -> Self {
         let vector_store = Arc::new(VectorStore::new(pool.clone()));
         
@@ -102,6 +173,7 @@ impl IngestWorker {
             vector_store,
             embeddings: Arc::new(embeddings),
             pipeline,
+            job_status_store,
         }
     }
 
@@ -128,13 +200,8 @@ impl IngestWorker {
             "Starting ingest job"
         );
 
-        // Update job status to processing
+        // Update job status to processing (via trait, testable)
-        if let Err(e) = sqlx::query("UPDATE ingest_jobs SET status=$1, started_at=NOW() WHERE ingest_id=$2")
+        if let Err(e) = self.job_status_store.update_status(ingest_id, JobStatus::Processing).await {
-            .bind(JobStatus::Processing.as_str())
-            .bind(ingest_id)
-            .execute(&self.pool)
-            .await
-        {
             tracing::error!(
                 target: "ingest",
                 error = %e,
@@ -213,14 +280,9 @@ impl IngestWorker {
             }
         }
 
-        // Mark job complete
+        // Mark job complete (via trait, testable)
         let final_status = JobStatus::Done;
-        if let Err(e) = sqlx::query("UPDATE ingest_jobs SET status=$1, completed_at=NOW() WHERE ingest_id=$2")
+        if let Err(e) = self.job_status_store.update_status(ingest_id, final_status).await {
-            .bind(final_status.as_str())
-            .bind(ingest_id)
-            .execute(&self.pool)
-            .await
-        {
             tracing::error!(
                 target: "ingest",
                 error = %e,
@@ -316,6 +378,7 @@ async fn save_entity_with_logging(
 }
 
 /// Save entity to database via raw SQL (normally would use EntityRepo trait)  
+/// NOTE: async_trait requires manual implementation for non-trait functions
 async fn save_entity_to_db(pool: &PgPool, entity: &mem_core::entity::Entity) -> Result<()> {
     // Convert OffsetDateTime to PostgreSQL timestamp format
     let t_created_str = entity.t_created.to_string();

tests/integration_ingest_with_gw.rs

@@ -0,0 +1,333 @@
+//! Integration test: Full ingest + embedding flow with api-gw
+//!
+//! Tests:
+//! 1. POST /memory/ingest with sample records
+//! 2. Poll /memory/ingest/{id} until done
+//! 3. Log root causes of errors
+//!
+//! Features:
+//! - RAII guard for port-forward cleanup (fix for resource leak)
+//! - Enhanced error handling with resource cleanup
+//!
+//! Requires:
+//! - DATABASE_URL set (postgres)
+//! - LLM_ENDPOINT set (for embeddings)
+//! - Server running locally or started by test
+//!
+//! Usage:
+//! ```
+//! RUST_LOG=debug cargo test --test integration_ingest_with_gw -- --nocapture
+//! ```
+
+use std::env;
+use std::time::Duration;
+use tokio::time::sleep;
+use serde_json::json;
+use std::process::Child;
+
+/// RAII guard for port-forward cleanup — ensures process is killed even if test panics
+struct PortForwardGuard {
+    process: Option<Child>,
+}
+
+impl PortForwardGuard {
+    fn spawn(namespace: &str, service: &str, local_port: u16, remote_port: u16) -> std::io::Result<Self> {
+        let process = std::process::Command::new("kubectl")
+            .args(&["-n", namespace, "port-forward", &format!("svc/{}", service), &format!("{}:{}", local_port, remote_port)])
+            .spawn()?;
+        
+        Ok(PortForwardGuard {
+            process: Some(process),
+        })
+    }
+}
+
+impl Drop for PortForwardGuard {
+    fn drop(&mut self) {
+        if let Some(mut process) = self.process.take() {
+            let _ = process.kill();
+            let _ = process.wait();
+        }
+    }
+}
+
+#[tokio::test]
+#[ignore] // Run manually: cargo test --test integration_ingest_with_gw -- --ignored --nocapture
+async fn test_ingest_with_embeddings_and_logging() {
+    // Initialize tracing with DEBUG level to see all logs
+    let _ = tracing_subscriber::fmt()
+        .with_max_level(tracing::Level::DEBUG)
+        .with_writer(std::io::stderr)
+        .try_init();
+
+    let base_url = env::var("MEM_API_URL").unwrap_or_else(|_| "http://localhost:8080".to_string());
+    let api_key = env::var("MEM_API_KEY").unwrap_or_else(|_| "test-key".to_string());
+    let local_port: u16 = 9990;
+    const NAMESPACE: &str = "poimen";
+    const SERVICE: &str = "poimen-memory";
+
+    // Spawn port-forward with RAII guard — guaranteed cleanup
+    let _pf_guard = match PortForwardGuard::spawn(NAMESPACE, SERVICE, local_port, 8080) {
+        Ok(guard) => {
+            sleep(Duration::from_secs(2)).await;
+            println!("[TEST] ✓ Port-forward started");
+            guard
+        }
+        Err(e) => {
+            eprintln!("[ERROR] Failed to spawn port-forward: {}", e);
+            return;
+        }
+    };
+
+    let client = reqwest::Client::new();
+
+    // Sample ingest payload
+    let payload = json!({
+        "project": "test-project",
+        "records": [
+            {
+                "content": "Kubernetes is an open-source container orchestration platform. [[Docker]] [[Go]]",
+                "source": "wiki/kubernetes"
+            },
+            {
+                "content": "Docker is a containerization platform that makes it easier to build, ship, and run applications. [[Linux]] [[Container]]",
+                "source": "wiki/docker"
+            },
+            {
+                "content": "Go is a programming language designed at Google. [[Concurrency]] [[Static Typing]]",
+                "source": "wiki/go"
+            }
+        ]
+    });
+
+    println!("[TEST] Sending ingest request...");
+    tracing::info!(
+        target: "integration_test",
+        "Ingest payload: {}",
+        serde_json::to_string_pretty(&payload).unwrap()
+    );
+
+    // POST /memory/ingest
+    let response = match client
+        .post(&format!("{}/memory/ingest", base_url))
+        .header("Authorization", format!("Bearer {}", api_key))
+        .json(&payload)
+        .send()
+        .await
+    {
+        Ok(resp) => resp,
+        Err(e) => {
+            eprintln!("[ERROR] Failed to send ingest request: {}", e);
+            tracing::error!(
+                target: "integration_test",
+                error = %e,
+                "Failed to POST /memory/ingest"
+            );
+            panic!("Request failed: {}", e);
+        }
+    };
+
+    let status = response.status();
+    println!("[TEST] Ingest response status: {}", status);
+
+    let body_text = match response.text().await {
+        Ok(text) => text,
+        Err(e) => {
+            tracing::error!(target: "integration_test", error = %e, "Failed to read response body");
+            panic!("Failed to read response body: {}", e);
+        }
+    };
+
+    println!("[TEST] Response body:\n{}", body_text);
+
+    // Parse response
+    let resp_json: serde_json::Value = match serde_json::from_str(&body_text) {
+        Ok(j) => j,
+        Err(e) => {
+            tracing::error!(
+                target: "integration_test",
+                error = %e,
+                body = %body_text,
+                "Failed to parse JSON response"
+            );
+            panic!("Failed to parse JSON: {}", e);
+        }
+    };
+
+    let ingest_id = match resp_json["id"].as_str() {
+        Some(id) => id.to_string(),
+        None => {
+            tracing::error!(
+                target: "integration_test",
+                response = %serde_json::to_string_pretty(&resp_json).unwrap(),
+                "Missing 'id' in response"
+            );
+            panic!("Missing 'id' in response: {}", resp_json);
+        }
+    };
+
+    println!("[TEST] Ingest ID: {}", ingest_id);
+    tracing::info!(target: "integration_test", ingest_id = %ingest_id, "Ingest queued");
+
+    // Poll until complete or timeout
+    let max_polls = 60; // 10 minutes with 10s intervals
+    for poll_num in 1..=max_polls {
+        sleep(Duration::from_secs(10)).await;
+
+        println!(
+            "[TEST] Poll #{}/{}: Checking status of ingest {}",
+            poll_num, max_polls, ingest_id
+        );
+
+        let status_response = match client
+            .get(&format!("{}/memory/ingest/{}", base_url, ingest_id))
+            .header("Authorization", format!("Bearer {}", api_key))
+            .send()
+            .await
+        {
+            Ok(resp) => resp,
+            Err(e) => {
+                tracing::error!(
+                    target: "integration_test",
+                    error = %e,
+                    ingest_id = %ingest_id,
+                    poll = poll_num,
+                    "Failed to fetch status"
+                );
+                eprintln!("[ERROR] Failed to fetch status: {}", e);
+                sleep(Duration::from_secs(5)).await;
+                continue;
+            }
+        };
+
+        let status_text = match status_response.text().await {
+            Ok(text) => text,
+            Err(e) => {
+                tracing::error!(
+                    target: "integration_test",
+                    error = %e,
+                    ingest_id = %ingest_id,
+                    "Failed to read status response"
+                );
+                eprintln!("[ERROR] Failed to read status: {}", e);
+                continue;
+            }
+        };
+
+        let status_json: serde_json::Value = match serde_json::from_str(&status_text) {
+            Ok(j) => j,
+            Err(e) => {
+                tracing::error!(
+                    target: "integration_test",
+                    error = %e,
+                    body = %status_text,
+                    "Failed to parse status JSON"
+                );
+                eprintln!("[ERROR] Failed to parse status JSON: {}", e);
+                continue;
+            }
+        };
+
+        let status = status_json["status"].as_str().unwrap_or("unknown");
+        println!(
+            "[TEST] Poll #{}: status = {}",
+            poll_num, status
+        );
+
+        tracing::info!(
+            target: "integration_test",
+            ingest_id = %ingest_id,
+            poll = poll_num,
+            status = %status,
+            full_response = %serde_json::to_string_pretty(&status_json).unwrap(),
+            "Status check"
+        );
+
+        match status {
+            "done" => {
+                println!("[TEST] ✓ Ingest completed successfully!");
+                tracing::info!(target: "integration_test", "Ingest completed");
+
+                // Extract and log results
+                if let Some(results) = status_json.get("results") {
+                    println!("[TEST] Results:\n{}", serde_json::to_string_pretty(results).unwrap());
+                    tracing::info!(
+                        target: "integration_test",
+                        results = %serde_json::to_string_pretty(results).unwrap(),
+                        "Ingest results"
+                    );
+                }
+                return;
+            }
+            "failed" | "error" => {
+                let error_msg = status_json["error"].as_str().unwrap_or("unknown error");
+                println!("[TEST] ✗ Ingest FAILED: {}", error_msg);
+                tracing::error!(
+                    target: "integration_test",
+                    ingest_id = %ingest_id,
+                    error = %error_msg,
+                    full_response = %serde_json::to_string_pretty(&status_json).unwrap(),
+                    "Ingest failed"
+                );
+                panic!("Ingest failed: {}", error_msg);
+            }
+            "processing" | "queued" => {
+                // Continue polling
+                println!("[TEST] Still processing, poll again...");
+            }
+            _ => {
+                println!("[TEST] Unknown status: {}", status);
+                tracing::warn!(target: "integration_test", status = %status, "Unknown status");
+            }
+        }
+    }
+
+    // Timeout — _pf_guard will be dropped here, cleaning up port-forward
+    let msg = format!("Ingest did not complete after {} polls (timeout)", max_polls);
+    println!("[TEST] ✗ {}", msg);
+    tracing::error!(target: "integration_test", ingest_id = %ingest_id, "Ingest timeout");
+    panic!("{}", msg);
+}
+
+#[tokio::test]
+#[ignore]
+async fn test_ingest_endpoint_only() {
+    let _ = tracing_subscriber::fmt()
+        .with_max_level(tracing::Level::DEBUG)
+        .try_init();
+
+    let base_url = env::var("MEM_API_URL").unwrap_or_else(|_| "http://localhost:8080".to_string());
+    let api_key = env::var("MEM_API_KEY").unwrap_or_else(|_| "test-key".to_string());
+
+    let client = reqwest::Client::new();
+
+    let payload = json!({
+        "project": "test-project",
+        "records": [
+            {
+                "content": "Simple test record",
+                "source": "test"
+            }
+        ]
+    });
+
+    println!("[TEST] Testing /memory/ingest endpoint only");
+
+    let response = client
+        .post(&format!("{}/memory/ingest", base_url))
+        .header("Authorization", format!("Bearer {}", api_key))
+        .json(&payload)
+        .send()
+        .await
+        .expect("Failed to send request");
+
+    println!("[TEST] Status: {}", response.status());
+
+    let body = response.text().await.expect("Failed to read body");
+    println!("[TEST] Response: {}", body);
+
+    let json: serde_json::Value = serde_json::from_str(&body).expect("Invalid JSON");
+    println!("[TEST] Parsed: {}", serde_json::to_string_pretty(&json).unwrap());
+
+    assert!(json.get("id").is_some(), "Response should contain 'id'");
+}

tests/unit_ingest_logging.rs

@@ -0,0 +1,341 @@
+//! Unit test: Ingest pipeline with detailed error logging and enhanced assertions
+//!
+//! Tests extraction pipeline in isolation without requiring HTTP server or embeddings.
+//! Useful for debugging extraction errors.
+//!
+//! Features:
+//! - Verify extracted entity names (not just count)
+//! - Verify edge connections between entities
+//! - Detailed error logging
+//!
+//! Usage:
+//! ```
+//! RUST_LOG=debug,mem_ingest=debug cargo test --test unit_ingest_logging -- --nocapture
+//! ```
+
+#[cfg(test)]
+mod tests {
+    use mem_ingest::ingest_pipeline::{IngestPipeline, Episode};
+    use mem_ingest::entity_extractor::WikiLinkFallbackExtractor;
+    use mem_ingest::fact_extractor::SimpleFactExtractor;
+    use mem_ingest::contradiction_detector::ContradictionHandler;
+    use std::sync::Arc;
+
+    fn init_logging() {
+        let _ = tracing_subscriber::fmt()
+            .with_max_level(tracing::Level::DEBUG)
+            .with_writer(std::io::stderr)
+            .try_init();
+    }
+
+    #[tokio::test]
+    async fn test_wiki_link_extraction_with_entity_verification() {
+        init_logging();
+
+        println!("\n[TEST] Wiki link extraction with entity name verification\n");
+
+        let entity_extractor = Arc::new(WikiLinkFallbackExtractor);
+        let fact_extractor = Arc::new(SimpleFactExtractor);
+        let contradiction_detector = Arc::new(ContradictionHandler::default());
+
+        let pipeline = IngestPipeline::new(
+            entity_extractor,
+            fact_extractor,
+            contradiction_detector,
+        );
+
+        let episode = Episode {
+            id: "test-1".to_string(),
+            project_id: "test-project".to_string(),
+            text: "Kubernetes [[Docker]] is a [[Container]] orchestration platform. It works with [[Go]] programs."
+                .to_string(),
+            wiki_links: vec!["Docker".to_string(), "Container".to_string(), "Go".to_string()],
+        };
+
+        tracing::info!(
+            target: "test",
+            episode_id = %episode.id,
+            wiki_links = ?episode.wiki_links,
+            "Starting pipeline ingest"
+        );
+
+        match pipeline.ingest(&episode).await {
+            Ok(result) => {
+                tracing::info!(
+                    target: "test",
+                    entities = result.entities.len(),
+                    edges = result.edges.len(),
+                    reviews = result.reviews.len(),
+                    "Pipeline succeeded"
+                );
+
+                println!("✓ Extracted {} entities", result.entities.len());
+                for entity in &result.entities {
+                    println!("  - {} ({}): {}", entity.name, entity.entity_type.as_str(), entity.summary.as_deref().unwrap_or(""));
+                }
+
+                println!("✓ Extracted {} edges", result.edges.len());
+                for edge in &result.edges {
+                    println!("  - {} --[{}]--> {}", edge.source_entity_id, edge.relation_type, edge.target_entity_id);
+                }
+
+                // ENHANCED: Verify extracted entity names (not just count)
+                assert!(!result.entities.is_empty(), "Should extract at least one entity");
+                
+                let entity_names: Vec<&str> = result.entities.iter().map(|e| e.name.as_str()).collect();
+                println!("\nEntity names extracted: {:?}", entity_names);
+                
+                assert!(
+                    entity_names.iter().any(|&name| name.contains("Docker") || name.contains("docker")),
+                    "Should extract Docker entity"
+                );
+                assert!(
+                    entity_names.iter().any(|&name| name.contains("Container") || name.contains("container")),
+                    "Should extract Container entity"
+                );
+                assert!(
+                    entity_names.iter().any(|&name| name.contains("Go") || name.contains("go")),
+                    "Should extract Go entity"
+                );
+
+                // ENHANCED: Verify edges connect correct entity pairs
+                if !result.edges.is_empty() {
+                    println!("\nEdge connections:");
+                    for edge in &result.edges {
+                        println!("  {} → {}", edge.source_entity_id, edge.target_entity_id);
+                        
+                        // Verify both endpoints exist in entities
+                        let source_exists = result.entities.iter().any(|e| e.id == edge.source_entity_id);
+                        let target_exists = result.entities.iter().any(|e| e.id == edge.target_entity_id);
+                        
+                        assert!(source_exists, "Edge source entity {} must exist in extracted entities", edge.source_entity_id);
+                        assert!(target_exists, "Edge target entity {} must exist in extracted entities", edge.target_entity_id);
+                    }
+                }
+            }
+            Err(e) => {
+                tracing::error!(
+                    target: "test",
+                    error = %e,
+                    "Pipeline failed"
+                );
+                panic!("Pipeline failed: {}", e);
+            }
+        }
+    }
+
+    #[tokio::test]
+    async fn test_extraction_error_logging() {
+        init_logging();
+
+        println!("\n[TEST] Pipeline error handling with logging\n");
+
+        let entity_extractor = Arc::new(WikiLinkFallbackExtractor);
+        let fact_extractor = Arc::new(SimpleFactExtractor);
+        let contradiction_detector = Arc::new(ContradictionHandler::default());
+
+        let pipeline = IngestPipeline::new(
+            entity_extractor,
+            fact_extractor,
+            contradiction_detector,
+        );
+
+        // Episode with problematic content (empty, or only whitespace)
+        let episode = Episode {
+            id: "test-empty".to_string(),
+            project_id: "test-project".to_string(),
+            text: "".to_string(),
+            wiki_links: vec![],
+        };
+
+        tracing::info!(
+            target: "test",
+            episode_id = %episode.id,
+            text_len = episode.text.len(),
+            "Processing empty episode"
+        );
+
+        match pipeline.ingest(&episode).await {
+            Ok(result) => {
+                tracing::info!(
+                    target: "test",
+                    entities = result.entities.len(),
+                    edges = result.edges.len(),
+                    "Empty episode processed (no error expected)"
+                );
+                println!("✓ Empty episode handled gracefully");
+            }
+            Err(e) => {
+                tracing::error!(
+                    target: "test",
+                    error = %e,
+                    "Empty episode caused error"
+                );
+                // Empty is OK for some extractors
+                println!("⚠ Empty episode error (may be expected): {}", e);
+            }
+        }
+    }
+
+    #[tokio::test]
+    async fn test_multiple_records_with_entity_verification() {
+        init_logging();
+
+        println!("\n[TEST] Processing multiple records with entity name verification\n");
+
+        let entity_extractor = Arc::new(WikiLinkFallbackExtractor);
+        let fact_extractor = Arc::new(SimpleFactExtractor);
+        let contradiction_detector = Arc::new(ContradictionHandler::default());
+
+        let pipeline = IngestPipeline::new(
+            entity_extractor,
+            fact_extractor,
+            contradiction_detector,
+        );
+
+        let records = vec![
+            ("Kubernetes [[Docker]] is a container orchestrator", "wiki/k8s"),
+            ("Docker [[Linux]] containers enable microservices", "wiki/docker"),
+            ("", "wiki/empty"),
+            ("Go [[Concurrency]] is powerful for backend services", "wiki/go"),
+        ];
+
+        let mut success_count = 0;
+        let mut error_count = 0;
+        let mut all_extracted_entities = Vec::new();
+
+        for (idx, (text, source)) in records.iter().enumerate() {
+            let episode = Episode {
+                id: format!("record-{}", idx),
+                project_id: "test-project".to_string(),
+                text: text.to_string(),
+                wiki_links: vec![],
+            };
+
+            tracing::info!(
+                target: "test",
+                record_idx = idx,
+                source = source,
+                text_len = text.len(),
+                "Processing record"
+            );
+
+            match pipeline.ingest(&episode).await {
+                Ok(result) => {
+                    tracing::debug!(
+                        target: "test",
+                        record_idx = idx,
+                        entities = result.entities.len(),
+                        edges = result.edges.len(),
+                        "Record succeeded"
+                    );
+                    println!("  ✓ Record {}: {} entities, {} edges", idx, result.entities.len(), result.edges.len());
+                    
+                    // Collect entity names for batch verification
+                    for entity in &result.entities {
+                        all_extracted_entities.push(entity.name.clone());
+                    }
+                    
+                    success_count += 1;
+                }
+                Err(e) => {
+                    tracing::warn!(
+                        target: "test",
+                        record_idx = idx,
+                        error = %e,
+                        source = source,
+                        "Record failed"
+                    );
+                    println!("  ✗ Record {}: {}", idx, e);
+                    error_count += 1;
+                }
+            }
+        }
+
+        println!("\nSummary: {} success, {} errors", success_count, error_count);
+        println!("All extracted entities: {:?}", all_extracted_entities);
+
+        tracing::info!(
+            target: "test",
+            total_records = records.len(),
+            success = success_count,
+            errors = error_count,
+            "Batch processing complete"
+        );
+
+        // ENHANCED: Verify that expected entities were extracted across records
+        assert!(success_count > 0, "At least some records should succeed");
+        assert!(
+            all_extracted_entities.iter().any(|name| name.contains("Docker") || name.contains("docker")),
+            "Docker entity should be extracted from at least one record"
+        );
+        assert!(
+            all_extracted_entities.iter().any(|name| name.contains("Linux") || name.contains("linux")),
+            "Linux entity should be extracted from at least one record"
+        );
+        assert!(
+            all_extracted_entities.iter().any(|name| name.contains("Go") || name.contains("go")),
+            "Go entity should be extracted from at least one record"
+        );
+    }
+
+    #[tokio::test]
+    async fn test_entity_deduplication() {
+        init_logging();
+
+        println!("\n[TEST] Entity deduplication (same entity from multiple records)\n");
+
+        let entity_extractor = Arc::new(WikiLinkFallbackExtractor);
+        let fact_extractor = Arc::new(SimpleFactExtractor);
+        let contradiction_detector = Arc::new(ContradictionHandler::default());
+
+        let pipeline = IngestPipeline::new(
+            entity_extractor,
+            fact_extractor,
+            contradiction_detector,
+        );
+
+        // Two records with overlapping entity references
+        let episode1 = Episode {
+            id: "record-1".to_string(),
+            project_id: "test-project".to_string(),
+            text: "Kubernetes uses [[Docker]] containers".to_string(),
+            wiki_links: vec!["Docker".to_string()],
+        };
+
+        let episode2 = Episode {
+            id: "record-2".to_string(),
+            project_id: "test-project".to_string(),
+            text: "Docker is used by [[Kubernetes]]".to_string(),
+            wiki_links: vec!["Kubernetes".to_string()],
+        };
+
+        let mut all_entities = Vec::new();
+
+        for episode in &[episode1, episode2] {
+            match pipeline.ingest(episode).await {
+                Ok(result) => {
+                    all_entities.extend(result.entities);
+                }
+                Err(e) => {
+                    tracing::error!(target: "test", error = %e, "Failed to ingest");
+                }
+            }
+        }
+
+        println!("Total entities extracted: {}", all_entities.len());
+        for entity in &all_entities {
+            println!("  - {}", entity.name);
+        }
+
+        // Verify both Docker and Kubernetes were extracted
+        assert!(
+            all_entities.iter().any(|e| e.name.contains("Docker") || e.name.contains("docker")),
+            "Docker should be extracted"
+        );
+        assert!(
+            all_entities.iter().any(|e| e.name.contains("Kubernetes") || e.name.contains("kubernetes")),
+            "Kubernetes should be extracted"
+        );
+    }
+}