Enhance FAQ and path command configuration documentation

- Added a new section in the FAQ detailing how to run meshcore-bot on a Raspberry Pi Zero 2 W, including specific configuration settings to manage memory usage effectively.
- Updated the path command configuration documentation to clarify the default values and usage of `graph_startup_load_days` and `graph_capture_enabled`, improving user understanding of these settings.

docs/faq.md

@@ -27,3 +27,28 @@ Without `--upgrade`, the script does *not* update the service file (systemd/laun
 ### How can I generate a custom command reference for my bot users?
 
 See [Custom command reference website](command-reference-website.md): it explains how to use `generate_website.py` to build a single-page HTML from your config (with optional styles) and upload it to your site.
+
+## Hardware and performance
+
+### How do I run meshcore-bot on a Raspberry Pi Zero 2 W?
+
+The Pi Zero 2 W has 512 MB of RAM, which is enough to run the bot, but the Mesh Graph can grow large on a busy mesh. Add the following to the `[Path_Command]` section of your `config.ini` to keep memory usage in check:
+
+```ini
+[Path_Command]
+# Limit startup memory: only load edges seen in the last 7 days.
+# Edges older than this have near-zero path confidence anyway.
+graph_startup_load_days = 7
+
+# Evict edges from RAM after 7 days without a new observation.
+graph_edge_expiration_days = 7
+
+# Write graph updates in batches rather than on every packet.
+graph_write_strategy = batched
+
+# If you don't use the !path command at all, disable graph capture
+# entirely to eliminate the background thread and all graph overhead.
+# graph_capture_enabled = false
+```
+
+These settings do not affect path prediction accuracy: edges older than a few days carry negligible confidence due to the 48-hour recency half-life used by the scoring algorithm.

docs/path-command-config.md

@@ -201,8 +201,15 @@ These settings control how graph edges are stored in the database.
 
 **`graph_startup_load_days`** (days, 0 = load all)
 - Load only edges seen in last N days on startup
-- `0` = load all edges (recommended for development)
-- Default: `0`
+- `0` = load all edges (use on servers with ample RAM)
+- Default: `14` (set to `0` in `config.ini` to load all)
+
+**`graph_capture_enabled`** (boolean)
+- When `false`, no new edge data is collected from packets and the background
+  batch writer thread is not started — reducing CPU and RAM overhead
+- Edges already in the database are still used for path validation
+- Set to `false` on devices that don't use the path command
+- Default: `true`
 
 ## Preset Configurations
 

modules/mesh_graph.py

@@ -533,13 +533,24 @@ class MeshGraph:
 
         try:
             if is_new:
-                # Insert new edge
+                # Upsert new edge.
+                # Use INSERT ... ON CONFLICT DO UPDATE so that if the row already exists
+                # in the database (e.g. it was filtered out of the in-memory graph at
+                # startup by startup_load_days / edge_expiration_days, or written by a
+                # concurrent process), we merge rather than fail with UNIQUE constraint.
                 query = '''
                     INSERT INTO mesh_connections
                     (from_prefix, to_prefix, from_public_key, to_public_key,
                      observation_count, first_seen, last_seen, avg_hop_position,
                      geographic_distance)
                     VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?)
+                    ON CONFLICT(from_prefix, to_prefix) DO UPDATE SET
+                        observation_count  = MAX(observation_count, excluded.observation_count),
+                        last_seen          = MAX(last_seen, excluded.last_seen),
+                        avg_hop_position   = excluded.avg_hop_position,
+                        geographic_distance = COALESCE(excluded.geographic_distance, geographic_distance),
+                        from_public_key    = COALESCE(excluded.from_public_key, from_public_key),
+                        to_public_key      = COALESCE(excluded.to_public_key, to_public_key)
                 '''
                 params = (
                     edge['from_prefix'],