Improve tailoring and job fetch resilience

README.md

@@ -59,6 +59,8 @@ RESUME_AGENT_API_KEY=your-rotated-token
 RESUME_AGENT_BASE_URL=https://service.nokod.ir/v1
 RESUME_AGENT_API_STYLE=chat
 RESUME_AGENT_LLM_TIMEOUT_SECONDS=1800
+RESUME_AGENT_JOB_FETCH_TIMEOUT_SECONDS=120
+RESUME_AGENT_JOB_FETCH_RETRIES=2
 ```
 
 Discover the provider's exact model IDs, then add your choice to `.env`:
@@ -87,6 +89,13 @@ RESUME_AGENT_LOG_MODEL_PAYLOADS=false
 to 1800 seconds (30 minutes), while connection establishment is capped at 30 seconds.
 An upstream gateway may still enforce its own shorter timeout.
 
+Fetching a job URL is a separate network request made before the model runs. It defaults
+to a 120-second timeout with two retries and exponential backoff. Configure it with
+`RESUME_AGENT_JOB_FETCH_TIMEOUT_SECONDS`, `RESUME_AGENT_JOB_FETCH_RETRIES`, and
+`RESUME_AGENT_JOB_FETCH_RETRY_BACKOFF_SECONDS`. If a careers site blocks server-side
+requests, paste the job description into the app instead; increasing the model timeout
+will not affect that connection.
+
 ## Usage
 
 ### Web interface