autocomplete

Richeduserus

I’ve followed the steps in this thread to set up the Autocomplete feature, but it's still not working for me.

Still, the Autocomplete hack isn’t completing my messages or giving suggestions. Everything else seems to work fine. No error messages appear either, so I’m not sure what’s missing.

Alexander01998

Hey Richeduserus,

Indeed, AutoComplete doesn't show any error message in chat when something is wrong, but it does show this error message in the logs. Check your log file (either latest.log in your .minecraft folder or the "show logs" option in your launcher).

This should tell you what is causing the issue. For example, it could be that your API key is invalid, or it could be that the model doesn't exist or that you don't have access to it, or that your OpenAI/OpenRouter account ran out of funds, or that your API endpoint is invalid, or that the servers are down. It could be a lot of different things.

Once you find the error message in your log file, if you find that you're having trouble interpreting it, feel free to post the error message here and I'll help you figure out what it means. In that case, please also post your AutoComplete settings, as they'll probably be important to know for troubleshooting.

Richeduserus

Alexander01998

Error:

$$
lessCopy code[11:14:46] [AutoComplete API Call/INFO]: [STDERR]: java.io.IOException: Server returned HTTP response code: 401 for URL: https://api.openai.com/v1/completions
$$

My settings:

$$
swiftCopy code{"stop":"\n","max_tokens":27,"temperature":1.22,"top_p":1.0,"presence_penalty":0.0,"frequency_penalty":0.0,"n":3,"model":"gpt-3.5-turbo-instruct","prompt":"=== Minecraft chat log ===\n<System>  iamnullman: hi\n<System> Cheating is against the server rules.\n<System> You can report rule breakers at reports.minecraft.net\n<System> ... "}
$$

Could you help me fix this?

Alexander01998

Richeduserus Okay, so we can see that the OpenAI server rejected your request with a status code of 401 (Unauthorized). This usually means that your API key is invalid. (relevant OpenAI docs)

There should be more details in the log entries just above or below that message, probably with a message that looks like this:

[11:14:46] [AutoComplete API Call/INFO]: [STDOUT]: {"error": {"message": "Incorrect API key provided: [example invalid key]. You can find your API key at https://platform.openai.com/account/api-keys.", "type": "invalid_request_error", "param": None, "code": "invalid_api_key"}}

To fix this, you will need to fix your WURST_OPENAI_KEY environment variable so that it matches your actual API key. Make sure there are no spaces before or after the key. It should start with sk-proj-, just like when you copied it from the OpenAI website.

Richeduserus

Alexander01998

I searched for terms like "incorrect", “fail,” “invalid,” and “error,” but didn’t find anything like that in the logs. The API key I’m using is the actual one from my OpenAI account — it starts with sk-proj-, and there are no extra spaces. Now I’m getting 429s instead. Looks like I’m being rate-limited — maybe too many requests? Not sure what exactly is triggering it.

Alexander01998

Richeduserus It seems like you're viewing the logs in your launcher, so make sure it's set to show all log entries and not just errors. The other message always appears, but its log level is set to INFO instead of ERROR.

Yes, 429 means you're being rate limited for sending too many requests (can also happen if you're using a VPN, meaning that other people on the same VPN IP are sending lots of requests).

Richeduserus

Hey Alexander01998,

I’ve set my API key and checked the logs, but AutoComplete still isn’t working. I also looked at my usage on the OpenAI dashboard — it says I have no credit and haven’t made any requests.

Could this be because I don’t have billing set up?
If so, are there any free options I could use with the AutoComplete?

Alexander01998

Richeduserus Yes to both.

It's probably because you don't have billing set up. OpenAI's free tier is extremely limited. You can't use most of their models and the ones you can use have very low rate limits. I can't seem to find an up-to-date list of which models are currently available in free tier. Actually, after searching for that list, I'm not even sure if the free tier still exists or if free OpenAI accounts are completely useless now.

There are many free options out there that you can use with AutoComplete, since it works with any OpenAI-compatible API. You can self-host an AI model for free by using something like the Oobabooga WebUI, but depending on how powerful your computer is, you'll probably be limited to very small models, which usually aren't very good. There are also several AI hosting companies that currently have some form of free tier, but the list changes often. Most of them are listed on OpenRouter, which also gives you an OpenAI-compatible API to use these models.

Richeduserus

Alexander01998

Checking out on OpenRouter. Any recommendations for a free model that works well for the AutoComplete? A free model has already been included, but it still doesn’t seem to work.

In the log:

[STDOUT]: {"stop":"\n","max_tokens":100,"temperature":1.79,"top_p":1.0,"presence_penalty":-1.04,"frequency_penalty":-1.01,"n":4,"model":"meta-llama/llama-4-scout:free","messages":[{"role":"system","content":"Complete the following text. Reply only with the completion. You are not an assistant."},{"role":"user","content":"=== Minecraft chat log ===\n<System> \n<System> Games! \n<achi> "}]}

[AutoComplete API Call/INFO]: [STDERR]: net.wurstclient.util.json.JsonException: Use JsonReader.setStrictness(Strictness.LENIENT) to accept malformed JSON at line 1 column 12 path $

See google/gsonblob/main/Troubleshooting.md#malformed-json

[AutoComplete API Call/INFO]: [STDERR]: at knot//net.wurstclient.util.json.JsonUtils.parseConnection(JsonUtils.java:111)

[AutoComplete API Call/INFO]: [STDERR]: at knot//net.wurstclient.util.json.JsonUtils.parseConnectionToObject(JsonUtils.java:134)

[AutoComplete API Call/INFO]: [STDERR]: at knot//net.wurstclient.hacks.autocomplete.OpenAiMessageCompleter.requestCompletions(OpenAiMessageCompleter.java:106)

[AutoComplete API Call/INFO]: [STDERR]: at knot//net.wurstclient.hacks.autocomplete.MessageCompleter.completeChatMessage(MessageCompleter.java:44)

[AutoComplete API Call/INFO]: [STDERR]: at knot//net.wurstclient.hacks.AutoCompleteHack.lambda$onUpdate$2(AutoCompleteHack.java:128)

[AutoComplete API Call/INFO]: [STDERR]: at java.base/java.lang.Thread.run(Thread.java:1583)

[AutoComplete API Call/INFO]: [STDERR]: Caused by: com.google.gson.stream.MalformedJsonException: Use JsonReader.setStrictness(Strictness.LENIENT) to accept malformed JSON at line 1 column 12 path $

See google/gsonblob/main/Troubleshooting.md#malformed-json

[AutoComplete API Call/INFO]: [STDERR]: at knot//com.google.gson.stream.JsonReader.syntaxError(JsonReader.java:1754)

[19:11:16] [AutoComplete API Call/INFO]: [STDERR]: at knot//com.google.gson.stream.JsonReader.checkLenient(JsonReader.java:1550)

[19:11:16] [AutoComplete API Call/INFO]: [STDERR]: at knot//com.google.gson.stream.JsonReader.doPeek(JsonReader.java:625)

[19:11:16] [AutoComplete API Call/INFO]: [STDERR]: at knot//com.google.gson.stream.JsonReader.peek(JsonReader.java:507)

[19:11:16] [AutoComplete API Call/INFO]: [STDERR]: at knot//com.google.gson.JsonParser.parseReader(JsonParser.java:111)

[19:11:16] [AutoComplete API Call/INFO]: [STDERR]: at knot//net.wurstclient.util.json.JsonUtils.parseConnection(JsonUtils.java:106)

[AutoComplete API Call/INFO]: [STDERR]: ... 5 more

To update — a model mistralai/mistral-7b-instruct:free has hit its daily request limit. In the error:

"Rate limit exceeded: free-models-per-day. Add 10 credits to unlock 1000 free model requests per day"

Any advice or suggestions?