Update README and main.go to enhance session organization and reduce timeout duration

Author: lemon07r

README.md

@@ -8,6 +8,9 @@ A fast, concurrent benchmarking tool for measuring LLM API performance metrics a
 - **Concurrent Testing**: Benchmark all providers simultaneously with `--all` flag
 - **Real Metrics**: Measures End-to-End Latency, Time to First Token (TTFT), and Throughput
 - **Accurate Token Counting**: Uses tiktoken for precise token measurements
+- **Session-Based Organization**: Each test run creates its own timestamped folder with logs and results
+- **Markdown Reports**: Auto-generates performance summaries with leaderboards and failure analysis
+- **Timeout Protection**: 2-minute timeout prevents indefinite hangs on stuck providers
 
 ## Quick Start
 
@@ -73,6 +76,28 @@ cp example.env .env
 ./llm-api-speed --all
 ```
 
+## Output
+
+Each test run creates a session folder: `results/session-YYYYMMDD-HHMMSS/`
+
+```
+results/session-20251110-004615/
+├── logs/
+│   ├── nim-20251110-004615.log
+│   ├── novita-20251110-004615.log
+│   └── minimax-20251110-004615.log
+├── nim-20251110-004637.json
+├── novita-20251110-004640.json
+├── minimax-20251110-004642.json
+└── REPORT.md  # Performance summary with leaderboards
+```
+
+**REPORT.md** includes:
+- Summary statistics (success/failure counts)
+- Performance leaderboards (by throughput and TTFT)
+- Detailed metrics for all providers
+- Error details for failed tests
+
 ## Supported Providers
 
 - **generic** - OpenRouter (default) or any OpenAI-compatible API (use `--url` to override)

main.go

@@ -88,8 +88,8 @@ func testProviderMetrics(config ProviderConfig, tke *tiktoken.Tiktoken, wg *sync
 	var firstTokenTime time.Time
 	var fullResponseContent strings.Builder
 
-	// Add timeout context to prevent indefinite hangs (5 minutes)
-	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Minute)
+	// Add timeout context to prevent indefinite hangs (2 minutes)
+	ctx, cancel := context.WithTimeout(context.Background(), 2*time.Minute)
 	defer cancel()
 
 	stream, err := client.CreateChatCompletionStream(ctx, req)