Docker compose and expected final release

.gitignore

@@ -140,3 +140,5 @@ behaviours/Z_repro_pathfinder_issue
 game_plan.json
 custom_game_plan.json
 custom_game_plan.json
+scripts/PRD-build_push_and_run_dispatcher_container.bat
+scripts/PRD-build-push-and-run-compose.bat

README.md

@@ -20,3 +20,30 @@ The project also contains a `dispatcher` which based on values in a configured d
 dispatcher reads the database for behaviours and 1-off tasks and executes matching `behaviours`   
 * `conductorWK25.py` - assigns behaviours to ships based on the contents of a custom game plan file like or populates its own default game plan.
 * `behaviours\` is a folder of python classes that inherit `generic_behaviour.py`. They all execute and initialise in the same way, and can be run from command line, or automatically the dispatcher.  
+
+
+
+## Setup
+
+You need to set the following environment variables:
+
+If you wanted to use a .env file, you could use the following format:
+
+```bash
+ST_TOKEN=eyJh....
+ST_DB_HOST=spacetraders_db_instance
+ST_DB_PASSWORD=spacetraders_pass
+ST_DB_NAME=spacetraders
+ST_DB_PORT=5432
+POSTGRES_PASSWORD=spacetraders_pass 
+```
+
+## Deploy
+
+You can use docker-compose to deploy the application.
+
+```bash
+docker compose  -f "scripts\all_in_one_compose.yml" up -d 
+```
+
+This will host the UI on port 3000, the DB on port 5432, and execute the dispatcher application with whatever token is specified in the .env file.

behaviours/generic_behaviour.py

@@ -41,8 +41,6 @@ def __init__(
             self.logger.warning(
                 "you're passing the connection in again, shouldn't be doing that. "
             )
-        saved_data = json.load(open(config_file_name, "r+"))
-        token = None
         try:
             self.priority = int(self.behaviour_params.get("priority", 5))
         except ValueError:
@@ -51,13 +49,7 @@ def __init__(
                 "priority value is invalid, defaulting to 5. Value was %s",
                 self.behaviour_params.get("priority", 5),
             )
-        for agent in saved_data["agents"]:
-            if agent.get("username", "") == agent_name:
-                token = agent["token"]
-                break
-        if not token:
-            # register the user
-            pass
+        token = os.environ.get("ST_TOKEN", None)
         db_host = os.environ.get("ST_DB_HOST", None)
         db_port = os.environ.get("ST_DB_PORT", None)
         db_name = os.environ.get("ST_DB_NAME", None)
@@ -906,6 +898,7 @@ def go_and_buy(
         if not resp:
             self.st.view_my_self(True)
             self.st.ships_view_one(ship.name, True)
+            self.log_market_changes(ship.nav.waypoint_symbol)
             resp = self.purchase_what_you_can(
                 target_tradegood, min(max_to_buy, ship.cargo_space_remaining)
             )
@@ -1084,7 +1077,7 @@ def sleep_until_arrived(self):
     self = bhvr
     st = self.st
     ship = self.ship
-
+    bhvr.sell_all_cargo()
     current_system = st.systems_view_one(ship.nav.system_symbol)
     factions = st.list_factions()
     for faction in factions:

behaviours/trade_best_intrasolar.py

@@ -139,7 +139,6 @@ def get_best_trade(self):
     ship = f"{agent}-{ship_number}"
     behaviour_params = {
         "priority": 3,
-        "target_sys": "X1-DM68",
     }
 
     bhvr = TradeBestInSystem(agent, ship, behaviour_params or {})

blog/Version 2.1.5 (week 31&32).md

@@ -8,6 +8,7 @@ With Justin back and active (and already in first place), the focus is going to
 Node-V isn't doing anything at all, when I restarted the dispatcher I got a bunch of errors from it attempting to map to a system it's not in?  
 Turns out we need to purge the cache between resets. Maybe the clint should do that if the system's dates don't match up with the service?
 
+
 ## performance
 
 | stat              | Node U (exp) | node V (26-27) | node C (27-28) |  
@@ -22,3 +23,62 @@ Turns out we need to purge the cache between resets. Maybe the clint should do t
 | average delay     | 
 | CPH               | 
 | CPR               |
+
+
+## Stagnation
+
+It's really feeling like we've not made any progress for a while.  
+With that in mind, it might be good to draw a line under this client, fix some of the structural issues with the SDK, tag it v2.0 and start again.
+
+
+
+### So what worked?
+
+* The "trade best" doesn't compete with Justin but it's very good in principle. 
+* The "Chain Trade" technology worked very well.
+* Having generic behaviour methods for travel, refuel, buy, sell, extract_till_full 
+* The "contracts" and both "management" behviours worked well but sometimes had downtime.
+* straightforward "mine and go sell" works
+* The request throttler and ship-specific priorities 
+* The pathfinder did great except for assumption that it always had the latest data.
+* Putting the DB and and the UI in containers was great for resets.
+* The UI being publicly hosted was great
+
+### What didn't work?
+
+* Weekly reset and caching issues were a pain. If we want an "always in memory" approach, the SDK will need to be more robust about managing when market data is old.  
+* The UI being a seperate process meant exposing logs to the player and getting feedback from actions was a right pain.  
+* The UI being an "all in one window" ended up being harder to work with than I'd hoped. 
+* Mistakes made in the SDK were hard to fix (seperate tables for markets etc...)
+* A limitation of the old design was that it really was one-ship per client, controlled by the DB, with the potential of split behaviour.
+
+
+### What should our new client do?
+
+* We should try to be as stateless as possible to allow for quick "off then on" again. 
+* Haulers should have three priorities - strengthen markets, trade ABUNDANT/ HIGH, and then haul mining goods & fuel.
+* Have a unified system for picking what to do next, and a thread-safe mechanism for avoiding collisions - ships should be aware of each others actions, states, and behaviour.
+* Have a UI that works on mobile. 
+* Have a UI that lets you take direct control of a ship, and activate any of our "blocks" of code (Travel, refuel, extract, buy, sell, build etc...)
+* Have a UI that shows market changes over time.
+* Have a UI that renders a production chain.
+* Allow multi-token operation - our public UI should be for any player to control their ship.
+* work when the user isn't on the browser
+* Visualise ship movements / distance in real time.
+* Have a comparison mode for previous resets.
+#### Archive the current client 
+
+* Get the dispatcher container up to speed, it should have a sigterm intercept for graceful shutdowns.
+  * the sigterm signal is being captured
+  * the compose file is written
+  * [the DB container is out of date]
+  * The UI container is able to connect to the DB
+  * the dispatcher is able to connect to the DB
+* Have a cron job that spins up a new container every 3 hours
+
+
+#### Development milestone 1
+
+* Token management from the UI.
+* Logs fedback to the user (websocket?)
+* execute move, buy, sell blocks to ships from the UI.

dispatcher.dockerfile

@@ -5,8 +5,11 @@ copy ./requirements.txt ./requirements.txt
 run python -m pip install -r requirements.txt
 
 workdir spacetraders_sdk
+
+
 run chmod +x ./setup.sh
-run /spacetraders_sdk/setup.sh
+run python3 -m build
+run python3 -m pip install dist/straders-2.1.4-py3-none-any.whl --force-reinstall
 
 workdir ..
 
@@ -15,4 +18,5 @@ copy . .
 ENV ST_DB_USER=spacetraders 
 
 
-CMD python dispatcherWK16.py 
+CMD python dispatcherWK25.py 
+