Gork bot

.cspell.json

@@ -27,13 +27,18 @@
     "livecrawl",
     "grok",
     "gork",
+    "imgork",
     "dalle",
     "dall",
     "arcas",
     "techwithanirudh",
     "selfbot",
     "shitposter",
     "cynicist",
-    "autobotting"
+    "autobotting",
+    "dotproduct",
+    "rerankinig",
+    "textembbeding",
+    "lefthook"
   ]
 }

.env.example

@@ -20,6 +20,7 @@ DISCORD_DEFAULT_GUILD_ID=your_discord_default_guild_id_here
 OPENAI_API_KEY=your_openai_api_key_here
 # HACKCLUB_API_KEY=your_hackclub_api_key_here
 # OPENROUTER_API_KEY=your_openrouter_api_key_here
+# COHERE_API_KEY=your_cohere_api_key_here
 
 # ---------------------------------------------------------------------------------------------------------
 # Redis

Dockerfile

@@ -0,0 +1,38 @@
+# use the official Bun image
+# see all versions at https://hub.docker.com/r/oven/bun/tags
+FROM oven/bun:1 AS base
+WORKDIR /usr/src/app
+
+RUN apt-get update \
+  && apt-get install -y --no-install-recommends \
+       curl git sudo python3 python3-pip make \
+  && rm -rf /var/lib/apt/lists/*
+
+# install dependencies into temp directory
+# this will cache them and speed up future builds
+FROM base AS install
+RUN mkdir -p /temp/dev
+COPY package.json bun.lock /temp/dev/
+RUN cd /temp/dev && bun install --frozen-lockfile
+
+# install with --production (exclude devDependencies)
+RUN mkdir -p /temp/prod
+COPY package.json bun.lock /temp/prod/
+RUN cd /temp/prod && bun install --production --ignore-scripts --frozen-lockfile
+
+# copy node_modules from temp directory
+# then copy all (non-ignored) project files into the image
+FROM base AS prerelease
+COPY --from=install /temp/dev/node_modules node_modules
+COPY . .
+
+# copy production dependencies and source code into final image
+FROM base AS release
+COPY --from=install /temp/prod/node_modules node_modules
+COPY --from=prerelease /usr/src/app . 
+RUN mkdir -p logs \
+ && chmod 0777 logs
+
+# run the app
+USER bun
+ENTRYPOINT [ "bun", "run", "start" ]

README.md

@@ -6,7 +6,7 @@
 2. 🚀 [Tech Stack](#tech-stack)
 3. 📚 [Getting Started](#getting-started)
 4. 🧠 [Memory](#memory)
-4. 📝 [License](#license)
+5. 📝 [License](#license)
 
 ## <a name="introduction">🤖 Introduction</a>
 
@@ -55,6 +55,7 @@ Set the `PINECONE_API_KEY` environment variable to your Pinecone API key.
 Then, create a Pinecone index and set the `PINECONE_INDEX` environment variable to the name of your Pinecone index.
 
 Spec:
+
 - Pinecone index should be dense
 - Dimension: 1536
 - Metric: dotproduct

TODO.md

@@ -28,8 +28,42 @@ Add context to when the bot is triggered—for example, whether it’s due to a
 
 Switch from Mem0 (free, limited plan) to a more efficient memory system like Pinecone or another vector store. Implement a better memory workflow with both long-term and short-term memory. This way, the bot can retain conversation history, summarize previous messages, and maintain context over time.
 
-Look into CrewAI or build your own custom memory system (a custom approach is likely more flexible). The goal is for Zenix to be more tightly integrated with both voice chat and text messages.
-
-Zenix should have unified memory per user across all servers—not separate memories per server. That way, the bot always remembers the same person no matter where they interact with it.
 Fix commands (broken on autobotting)
-Cleanup memory part later
+Cleanup memory part later
+
+use lefthook instead of husky
+
+Add tool calling to memory, also use a FIFO queue instead of async sending and calculate WPM + ai response assumptions
+Properly refactor the memory system with querying like B does it
+Cleanup the code a bit
+Implement the BM25 thing
+give llm choice to reply or to generally not
+
+When pinging users mention @username then convert it to user ID like frank
+Improve system prompt to include tools
+When there is an attachment which it cant see add Right now it just adds [Attachments: png, zip, png, png] for each file attached
+when if it is not a type it can process, remove all blank stuff messages (https://github.com/DomBom16/frank/blob/main/src/utils/aiResponse.ts)
+convert status or whatever in discord the yellow white block to text like frank
+Also another issue is the build context thing's reference replies are broken
+
+Refactor the ping system, the bot pings random people and doesn't know who is who
+Add edit delete messages functionality for gork, meaning it can understand when messages were edited and also edit it's own messages liek Frank
+
+Refactor logging system with child loggers
+
+Refactor the adding metadata so guild and channel are not strings, and are JSON but the retrieval converts JSON to string and vice versa
+
+Implement text management logic like https://github.com/Giantpizzahead/bob-bot/blob/6574d0e988d6249c8df2a72179503e0d16f95a3c/src/bobbot/discord_helpers/text_channel_history.py#L127
+Do not INGEST the searchMemories tool when using addTool
+Give the AI Ability to use custom filters to searchMemories through tool calling, and ask it to use the info in the first searchMemories calls to call the second one, say that serve rnames etc are very important also give it what metadta params it has
+Give it the ability to raise a feature request which opens a jira ticket
+Have a custom memory manager like text channel history
+Improve logging to add a debug mode or reduce excessive loggin
+
+The bot has a habit of not replying to the designated message
+Input what activity the bot is performing, and add more set of activites
+Add debug tools like bob
+Add a agent before tasks which checks if it is a command or prompt inj
+Add a database (drizzle) with statistics on chat with servers
+
+Have a mode called selfbot and real bot, selfbot doesn't have commands etc, newinfra wow

VECTOR.md

@@ -1,18 +1,23 @@
-# Vector
-Let's use Pinecone for the following:
-1. As we use the AI SDK, we will use OpenAI compatible models for embedding, like embedding large (use models, embedding in models.ts), same with searching for data.
-2. We use a dense index in Pinecone for semantic search, matching the output of modern embedding models and the bot's use case. (https://docs.pinecone.io/guides/index-data/create-an-index#bring-your-own-vectors), include metdata like timestamp, user, server, channel, etc. So when searching we can limit the search to a specific server, channel, etc. (https://docs.pinecone.io/guides/search/filter-by-metadata#search-with-a-metadata-filter)
-3. We do NOT need to chunk the data as discord messages are already short and concise.
-4. When a message is deleted or edited, we should remove the embedding from the index, or update it. (https://docs.pinecone.io/guides/index-data/data-modeling#delete-chunks, https://docs.pinecone.io/guides/index-data/data-modeling#update-chunks). Have a way to easily connect message and embedding id.
-5. After all this we can upsert the embeddings to the index. (https://docs.pinecone.io/guides/index-data/upsert-data#upsert-dense-vectors) (https://docs.pinecone.io/guides/index-data/indexing-overview#bring-your-own-vectors)
-6. Setup a hybrid search index (https://docs.pinecone.io/guides/search/hybrid-search#use-a-single-hybrid-index)
-
-# Searches
-- When searching have a sophisticated algo (bob), check through server, recent messages, and a lot of context aswell so we get the data. w/metadata timeframes, users etc. (https://docs.pinecone.io/guides/get-started/quickstart#5-semantic-search)
-
-
-# TODO 
-Learn about namespaces
-Learn about rerankinig (https://docs.pinecone.io/guides/get-started/quickstart#6-rerank-results)
-Look at bob code (https://docs.pinecone.io/guides/get-started/quickstart#7-improve-results)
-Look at AI SDK pinecone example so we can implement it
+# Vector
+
+Let's use Pinecone for the following:
+
+1. As we use the AI SDK, we will use OpenAI compatible models for embedding, like embedding large (use models, embedding in models.ts), same with searching for data.
+2. We use a dense index in Pinecone for semantic search, matching the output of modern embedding models and the bot's use case. (https://docs.pinecone.io/guides/index-data/create-an-index#bring-your-own-vectors), include metadata like timestamp, user, server, channel, etc. So when searching we can limit the search to a specific server, channel, etc. (https://docs.pinecone.io/guides/search/filter-by-metadata#search-with-a-metadata-filter)
+3. We do NOT need to chunk the data as discord messages are already short and concise.
+4. When a message is deleted or edited, we should remove the embedding from the index, or update it. (https://docs.pinecone.io/guides/index-data/data-modeling#delete-chunks, https://docs.pinecone.io/guides/index-data/data-modeling#update-chunks). Have a way to easily connect message and embedding id.
+5. After all this we can upsert the embeddings to the index. (https://docs.pinecone.io/guides/index-data/upsert-data#upsert-dense-vectors) (https://docs.pinecone.io/guides/index-data/indexing-overview#bring-your-own-vectors)
+6. Setup a hybrid search index (https://docs.pinecone.io/guides/search/hybrid-search#use-a-single-hybrid-index)
+
+# Searches
+
+- When searching have a sophisticated algo (bob), check through server, recent messages, and a lot of context as well so we get the data. w/metadata timeframes, users etc. (https://docs.pinecone.io/guides/get-started/quickstart#5-semantic-search)
+
+# TODO
+
+Learn about namespaces
+Learn about rerankinig (https://docs.pinecone.io/guides/get-started/quickstart#6-rerank-results)
+Look at bob code (https://docs.pinecone.io/guides/get-started/quickstart#7-improve-results)
+Look at AI SDK pinecone example so we can implement it
+
+3-large-textembbeding