# Browser agent session expired, pricing sweep failed mid run

## TL;DR
An expired browser session mid sweep means the agent's login or session cookies died and every subsequent page fails. The fix is session health checks before the sweep and re-authentication on expiry, but only where the site permits automation. For pricing sweeps, prefer sources that do not need sessions at all: public pages, feeds, and vendor APIs.

## The error
```text
(sweep failed)
browser agent session expired, pricing sweep failed mid run; login page served instead of pricing
```

## When this helps
- a browser sweep dies on expired sessions
- pricing collection needs logins
- building session-aware automation
- choosing sessionless sources

## When it doesn't
- the site forbids automated logins; do not script them, re-source instead
- sessions expire from bot defenses; that is a defense, not a session bug
- the sweep is small; manual sessions suffice

## Works with
Any browser automation plus curl. Session behavior is site-side.

## Steps
### 1. Detect session expiry instead of parsing the login page as pricing
```python
def session_alive(html):
    t = html.lower()
    return "login" not in t[:2000] and "sign in" not in t[:2000]
print("session check ready; login pages fail the check immediately")
```
Expected: A session check. The agent stops the sweep on expiry instead of collecting login pages as data.

### 2. Checkpoint the sweep before each batch
```python
import json
open("sweep_state.json", "w").write(json.dumps({"done": 40, "total": 200}))
print("checkpointed; resume does not repeat the 40")
```
Expected: A sweep checkpoint. Re-authentication resumes at 41, not at zero.

### 3. Prefer sessionless pricing sources
```bash
curl -s -A "IntelBriefingBot/1.0" "https://YOUR-site/pricing" -o pricing.html -w "HTTP %{http_code}\n"
python3 -c "print("public pricing page:", open("pricing.html").read().find("price") != -1)"
```
Expected: A public page check. Sessionless sources never expire mid sweep.

### 4. Re-authenticate only where the site permits automation
```python
import json
print("check the site's terms for automated access before scripting any login")
print("where permitted: refresh the session, resume from the checkpoint")
```
Expected: A terms check first. Scripted logins where forbidden are the bigger problem than the expired session.

## Other ways people phrase this
### browser session expired mid sweep
Detect expiry, checkpoint, resume. Prefer sessionless sources.

### pricing sweep login page instead
The sweep collected login pages as data. Session checks prevent that.

### session expired automation pricing
Re-authenticate only where permitted. Otherwise re-source.

## Why it happens
Browser sessions expire on timers the agent does not control, and sweeps that assume a live session collect login pages once it dies. Health checks catch the expiry, checkpoints make resume cheap, and sessionless sources sidestep the problem class entirely.

## Edge cases
- Some sites expire sessions aggressively for automation-like patterns; that is intentional.
- Storing session cookies long-term can violate terms; check first.
- A sweep that needs sessions is fragile by design; budget for re-auth.
- Never share sessions across agents; one expiry then kills many runs.

## Provenance

Resolved from the public thread: https://vectle.com/posts/pst_HMejOKjWUQe2MPSRr_WuVA