feat: Structured Log Store + Interactive Log Explorer

README.md

@@ -88,6 +88,8 @@ uv run logan view -d "tmp/output"
 # server should be available at http://localhost:8000/log_diagnosis
 ``` 
 
+![Log-Explorer](./docs/asset/log-explorer.png)
+
 ## MCP Server (for AI Agents)
 
 LogAn exposes its analysis capabilities via the [Model Context Protocol (MCP)](https://modelcontextprotocol.io), allowing AI agents (Claude Desktop, Claude Code, custom agents) to analyze logs programmatically.

logan/cli.py

@@ -261,8 +261,8 @@ def analyze(files, glob, time_range, output_dir, debug_mode, process_all_files,
     click.echo(click.style("\n" + "=" * 50, fg="green"))
     click.echo(click.style("Analysis complete!", fg="green", bold=True))
     click.echo(f"\nOutput files:")
-    click.echo(f"  Anomaly report: {os.path.join(output_dir, 'log_diagnosis', 'anomalies.html')}")
-    click.echo(f"  Summary report: {os.path.join(output_dir, 'log_diagnosis', 'summary.html')}")
+    click.echo(f"  Log explorer:    {os.path.join(output_dir, 'log_diagnosis', 'explorer.html')}")
+    click.echo(f"  Store (Parquet): {os.path.join(output_dir, 'store', '')}")
 
 
 @cli.command()

logan/drain/run_drain.py

@@ -7,6 +7,8 @@
 from drain3.template_miner_config import TemplateMinerConfig
 from drain3.file_persistence import FilePersistence
 
+from logan.store.store import LogStore
+
 class Templatizer:
     """
     The Templatizer class is responsible for mining log templates using the DRAIN3 algorithm.
@@ -91,15 +93,26 @@ def miner(self, df, output_dir: str, template: str):
         # Initialize the TemplateMiner with the loaded configuration and file persistence
         template_miner_temporary = TemplateMiner(mem_persistence, config)
 
+        # Preserve original log text before Drain3 masking overwrites it
+        df["original_text"] = df["text"].astype(str)
+
         # Initialize a dictionary to store the loglines grouped by template IDs
         template_log_dict = {}
 
         # Mine templates by iterating directly over the column (avoids pd.Series overhead of df.apply)
         try:
             test_ids = []
+            template_strs = []
+            variables_list = []
             for log in df["truncated_log"].values:
-                test_ids.append(template_miner_temporary.add_log_message(log)['cluster_id'])
+                result = template_miner_temporary.add_log_message(log)
+                test_ids.append(result['cluster_id'])
+                tmpl = result.get('template_mined', '')
+                template_strs.append(tmpl)
+                variables_list.append(LogStore.extract_variables(log, tmpl))
             df["test_ids"] = test_ids
+            df["template_str"] = template_strs
+            df["variables"] = [json.dumps(v) for v in variables_list]
 
             if (self.debug_mode == "true"):
                 template_log_dict = df.groupby("test_ids")["truncated_log"].agg(list).to_dict()

logan/log_diagnosis/anomaly.py

@@ -2,11 +2,11 @@
 import json
 import pandas as pd
 import time
-import csv
 from .core import Core
 from datetime import datetime
-from logan.log_diagnosis.utils import get_anomaly_html_str, get_summary_html_str, compute_golden_signal_timeline
+from logan.log_diagnosis.utils import get_explorer_html_str, compute_golden_signal_timeline
 from logan.log_diagnosis.models import ModelManager, AllModels, ModelType
+from logan.store.store import LogStore
 
 class Anomaly(Core):
     """
@@ -307,21 +307,16 @@ def get_anomaly_report(self, df_inference_csv, output_dir):
             'list_templates': template_ids2
         })
 
-        # Read debug information (ignored and processed files)
-        with open(os.path.join(developer_debug_dir, "ignored_files.log"), 'r') as reader:
-            ignored_files = reader.read().splitlines()
-
-        with open(os.path.join(developer_debug_dir, "processed_files.log"), 'r') as reader:
-            processsed_files = reader.read().splitlines()
-
-        # Generate the HTML table for the anomaly report
-        html_table = get_anomaly_html_str(df_final_anomalies, output_dir)
-        with open(os.path.join(log_diagnosis_dir, "anomalies.html"), "w") as f:
-            f.write(html_table)
-
-        # Generate the HTML table for the summary report
-        html_table = get_summary_html_str(df_for_summary_html, include_golden_signal_dropdown=True, ignored_file_list=ignored_files, processed_file_list=processsed_files, output_dir=output_dir, has_timeline_data=has_timeline_data)
-        with open(os.path.join(log_diagnosis_dir, "summary.html"), "w") as f:
-            f.write(html_table)
+        # Build and persist the structured log store, then render explorer
+        try:
+            store = LogStore(output_dir)
+            store.build_from_df(df_inference_csv, temp_id_to_signal_map)
+            store.save_parquet()
+            store_meta = store.save_json_for_explorer()
+            html_explorer = get_explorer_html_str(store_meta)
+            with open(os.path.join(log_diagnosis_dir, "explorer.html"), "w") as f:
+                f.write(html_explorer)
+        except Exception as exc:
+            print(f"Warning: could not generate log store / explorer: {exc}")
 
         self.compute_anomaly_statistics(output_dir, (time.time() - start) * 1000)