aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorterminaldweller <devi@terminaldweller.com>2024-10-28 22:28:16 +0000
committerterminaldweller <devi@terminaldweller.com>2024-10-28 22:28:16 +0000
commite22d58cee6d9bb963f879fde8890e7743269afb3 (patch)
tree7521268ce29b4fca3b97bfe7451a828b7b880ca5
parentadded a new option, context. fixed a bug with the custom commands where the c... (diff)
downloadmilla-e22d58cee6d9bb963f879fde8890e7743269afb3.tar.gz
milla-e22d58cee6d9bb963f879fde8890e7743269afb3.zip
added openrouter as a provider
-rw-r--r--README.md4
-rw-r--r--main.go38
-rw-r--r--makefile30
-rw-r--r--openrouter.go200
-rw-r--r--plugins.go16
-rw-r--r--types.go31
6 files changed, 315 insertions, 4 deletions
diff --git a/README.md b/README.md
index 069ef7e..b6fa514 100644
--- a/README.md
+++ b/README.md
@@ -2,7 +2,7 @@
Milla is an IRC bot that:
-- sends things over to an LLM when you ask it questions and prints the answer with optional syntax-highlighting.Currently supported providers: Ollama, Openai, Gemini <br/>
+- sends things over to an LLM when you ask it questions and prints the answer with optional syntax-highlighting.Currently supported providers: Ollama, Openai, Gemini, Openrouter <br/>
- Milla can run more than one instance of itself
- Each instance can connect to a different ircd, and will get the full set of configs, e.g. different proxies, different postgres instance, ...
- You can define custom commands in the form of SQL queries to the database with the SQL query result being passed to the bot along with the given prompt and an optional limit so you don't go bankrupt(unless you are running ollama locally like the smart cookie that you are).<br/>
@@ -45,7 +45,7 @@ The SASL username.
The SASL password for SASL plain authentication. Can also be passed as and environment variable.
-#### ollamaEndpoint
+#### Endpoint
The address for the Ollama chat endpoint.
diff --git a/main.go b/main.go
index 4999c9a..e2f7c7b 100644
--- a/main.go
+++ b/main.go
@@ -405,6 +405,27 @@ func handleCustomCommand(
if result != "" {
sendToIRC(client, event, result, appConfig.ChromaFormatter)
}
+ case "openrouter":
+ var memory []MemoryElement
+
+ for _, log := range logs {
+ memory = append(memory, MemoryElement{
+ Role: "user",
+ Content: log.Log,
+ })
+ }
+
+ for _, customContext := range customCommand.Context {
+ memory = append(memory, MemoryElement{
+ Role: "user",
+ Content: customContext,
+ })
+ }
+
+ result := ORRequestProcessor(appConfig, client, event, &memory, customCommand.Prompt)
+ if result != "" {
+ sendToIRC(client, event, result, appConfig.ChromaFormatter)
+ }
default:
}
}
@@ -681,7 +702,7 @@ func DoOllamaRequest(
ctx, cancel := context.WithTimeout(context.Background(), time.Duration(appConfig.RequestTimeout)*time.Second)
defer cancel()
- request, err := http.NewRequest(http.MethodPost, appConfig.OllamaEndpoint, bytes.NewBuffer(jsonPayload))
+ request, err := http.NewRequest(http.MethodPost, appConfig.Endpoint, bytes.NewBuffer(jsonPayload))
if err != nil {
return "", err
@@ -1011,6 +1032,10 @@ func DoChatGPTRequest(
config := openai.DefaultConfig(appConfig.Apikey)
config.HTTPClient = &httpClient
+ if appConfig.Endpoint != "" {
+ config.BaseURL = appConfig.Endpoint
+ log.Print(config.BaseURL)
+ }
gptClient := openai.NewClientWithConfig(config)
@@ -1264,6 +1289,8 @@ func runIRC(appConfig TomlConfig) {
var GPTMemory []openai.ChatCompletionMessage
+ var ORMemory []MemoryElement
+
poolChan := make(chan *pgxpool.Pool, 1)
irc := girc.New(girc.Config{
@@ -1363,6 +1390,15 @@ func runIRC(appConfig TomlConfig) {
}
ChatGPTHandler(irc, &appConfig, &GPTMemory)
+ case "openrouter":
+ for _, context := range appConfig.Context {
+ ORMemory = append(ORMemory, MemoryElement{
+ Role: "user",
+ Content: context,
+ })
+ }
+
+ ORHandler(irc, &appConfig, &ORMemory)
}
go LoadAllPlugins(&appConfig, irc)
diff --git a/makefile b/makefile
new file mode 100644
index 0000000..b891015
--- /dev/null
+++ b/makefile
@@ -0,0 +1,30 @@
+.PHONY: d_test d_deploy d_down d_build help
+
+IMAGE_NAME=milla
+
+d_test:
+ nq docker compose -f ./docker-compose-devi.yaml up --build
+
+d_deploy:
+ nq docker compose -f ./docker-compose.yaml up --build
+
+d_down:
+ docker compose -f ./docker-compose.yaml down
+ docker compose -f ./docker-compose-devi.yaml down
+
+d_build: d_build_distroless_vendored
+
+d_build_regular:
+ docker build -t $(IMAGE_NAME)-f ./Dockerfile .
+
+d_build_distroless:
+ docker build -t $(IMAGE_NAME) -f ./Dockerfile_distroless .
+
+d_build_distroless_vendored:
+ docker build -t $(IMAGE_NAME) -f ./Dockerfile_distroless_vendored .
+
+help:
+ @echo "d_test"
+ @echo "d_deploy"
+ @echo "d_down"
+ @echo "d_build"
diff --git a/openrouter.go b/openrouter.go
new file mode 100644
index 0000000..8a1a1e5
--- /dev/null
+++ b/openrouter.go
@@ -0,0 +1,200 @@
+package main
+
+import (
+ "bytes"
+ "context"
+ "encoding/json"
+ "log"
+ "net"
+ "net/http"
+ "net/url"
+ "strings"
+ "time"
+
+ "github.com/alecthomas/chroma/v2/quick"
+ "github.com/lrstanley/girc"
+ "golang.org/x/net/proxy"
+)
+
+func DoORRequest(
+ appConfig *TomlConfig,
+ memory *[]MemoryElement,
+ prompt string,
+) (string, error) {
+ var jsonPayload []byte
+
+ var err error
+
+ memoryElement := MemoryElement{
+ Role: "user",
+ Content: prompt,
+ }
+
+ if len(*memory) > appConfig.MemoryLimit {
+ *memory = []MemoryElement{}
+
+ for _, context := range appConfig.Context {
+ *memory = append(*memory, MemoryElement{
+ Role: "assistant",
+ Content: context,
+ })
+ }
+ }
+
+ *memory = append(*memory, memoryElement)
+
+ ollamaRequest := OllamaChatRequest{
+ Model: appConfig.Model,
+ Messages: *memory,
+ }
+
+ jsonPayload, err = json.Marshal(ollamaRequest)
+ if err != nil {
+
+ return "", err
+ }
+
+ log.Printf("json payload: %s", string(jsonPayload))
+
+ ctx, cancel := context.WithTimeout(context.Background(), time.Duration(appConfig.RequestTimeout)*time.Second)
+ defer cancel()
+
+ request, err := http.NewRequest(http.MethodPost, appConfig.Endpoint, bytes.NewBuffer(jsonPayload))
+ if err != nil {
+
+ return "", err
+ }
+
+ request = request.WithContext(ctx)
+ request.Header.Set("content-type", "application/json")
+ request.Header.Set("Authorization", "Bearer "+appConfig.Apikey)
+
+ var httpClient http.Client
+
+ var dialer proxy.Dialer
+
+ if appConfig.LLMProxy != "" {
+ proxyURL, err := url.Parse(appConfig.IRCProxy)
+ if err != nil {
+ cancel()
+
+ log.Fatal(err.Error())
+ }
+
+ dialer, err = proxy.FromURL(proxyURL, &net.Dialer{Timeout: time.Duration(appConfig.RequestTimeout) * time.Second})
+ if err != nil {
+ cancel()
+
+ log.Fatal(err.Error())
+ }
+
+ httpClient = http.Client{
+ Transport: &http.Transport{
+ Dial: dialer.Dial,
+ },
+ }
+ }
+ response, err := httpClient.Do(request)
+
+ if err != nil {
+ return "", err
+ }
+
+ defer response.Body.Close()
+
+ log.Println("response body:", response.Body)
+
+ var orresponse ORResponse
+
+ err = json.NewDecoder(response.Body).Decode(&orresponse)
+ if err != nil {
+ return "", err
+ }
+
+ var result string
+
+ for _, choice := range orresponse.Choices {
+ result += choice.Message.Content + "\n"
+ }
+
+ return result, nil
+}
+
+func ORRequestProcessor(
+ appConfig *TomlConfig,
+ client *girc.Client,
+ event girc.Event,
+ memory *[]MemoryElement,
+ prompt string,
+) string {
+ response, err := DoORRequest(appConfig, memory, prompt)
+ if err != nil {
+ client.Cmd.ReplyTo(event, "error: "+err.Error())
+
+ return ""
+ }
+
+ assistantElement := MemoryElement{
+ Role: "assistant",
+ Content: response,
+ }
+
+ *memory = append(*memory, assistantElement)
+
+ log.Println(response)
+
+ var writer bytes.Buffer
+
+ err = quick.Highlight(&writer,
+ response,
+ "markdown",
+ appConfig.ChromaFormatter,
+ appConfig.ChromaStyle)
+ if err != nil {
+ client.Cmd.ReplyTo(event, "error: "+err.Error())
+
+ return ""
+ }
+
+ return writer.String()
+}
+
+func ORHandler(
+ irc *girc.Client,
+ appConfig *TomlConfig,
+ memory *[]MemoryElement) {
+ irc.Handlers.AddBg(girc.PRIVMSG, func(client *girc.Client, event girc.Event) {
+ if !strings.HasPrefix(event.Last(), appConfig.IrcNick+": ") {
+ return
+ }
+
+ if appConfig.AdminOnly {
+ byAdmin := false
+
+ for _, admin := range appConfig.Admins {
+ if event.Source.Name == admin {
+ byAdmin = true
+ }
+ }
+
+ if !byAdmin {
+ return
+ }
+ }
+
+ prompt := strings.TrimPrefix(event.Last(), appConfig.IrcNick+": ")
+ log.Println(prompt)
+
+ if string(prompt[0]) == "/" {
+ runCommand(client, event, appConfig)
+
+ return
+ }
+
+ result := ORRequestProcessor(appConfig, client, event, memory, prompt)
+ if result != "" {
+ sendToIRC(client, event, result, appConfig.ChromaFormatter)
+ }
+ })
+
+}
diff --git a/plugins.go b/plugins.go
index 3e29570..0ffaba2 100644
--- a/plugins.go
+++ b/plugins.go
@@ -238,6 +238,21 @@ func ircPartChannelClosure(luaState *lua.LState, client *girc.Client) func(*lua.
}
}
+func orRequestClosure(luaState *lua.LState, appConfig *TomlConfig) func(*lua.LState) int {
+ return func(luaState *lua.LState) int {
+ prompt := luaState.CheckString(1)
+
+ result, err := DoORRequest(appConfig, &[]MemoryElement{}, prompt)
+ if err != nil {
+ LogError(err)
+ }
+
+ luaState.Push(lua.LString(result))
+
+ return 1
+ }
+}
+
func ollamaRequestClosure(luaState *lua.LState, appConfig *TomlConfig) func(*lua.LState) int {
return func(luaState *lua.LState) int {
prompt := luaState.CheckString(1)
@@ -334,6 +349,7 @@ func millaModuleLoaderClosure(luaState *lua.LState, client *girc.Client, appConf
"send_ollama_request": lua.LGFunction(ollamaRequestClosure(luaState, appConfig)),
"send_gemini_request": lua.LGFunction(geminiRequestClosure(luaState, appConfig)),
"send_chatgpt_request": lua.LGFunction(chatGPTRequestClosure(luaState, appConfig)),
+ "send_or_request": lua.LGFunction(orRequestClosure(luaState, appConfig)),
"query_db": lua.LGFunction(dbQueryClosure(luaState, appConfig)),
"register_cmd": lua.LGFunction(registerLuaCommand(luaState, appConfig)),
"url_encode": lua.LGFunction(urlEncode(luaState)),
diff --git a/types.go b/types.go
index 1fc09dc..b976840 100644
--- a/types.go
+++ b/types.go
@@ -56,7 +56,7 @@ type TomlConfig struct {
IrcNick string `toml:"ircNick"`
IrcSaslUser string `toml:"ircSaslUser"`
IrcSaslPass string `toml:"ircSaslPass"`
- OllamaEndpoint string `toml:"ollamaEndpoint"`
+ Endpoint string `toml:"endpoint"`
Model string `toml:"model"`
ChromaStyle string `toml:"chromaStyle"`
ChromaFormatter string `toml:"chromaFormatter"`
@@ -79,6 +79,7 @@ type TomlConfig struct {
WebIRCHostname string `toml:"webIRCHostname"`
WebIRCAddress string `toml:"webIRCAddress"`
RSSFile string `toml:"rssFile"`
+ AnthropicVersion string `toml:"anthropicVersion"`
Plugins []string `toml:"plugins"`
Context []string `toml:"context"`
CustomCommands map[string]CustomCommand `toml:"customCommands"`
@@ -176,6 +177,34 @@ type OllamaChatRequest struct {
Messages []MemoryElement `json:"messages"`
}
+type ORMessage struct {
+ Role string `json:"role"`
+ Content string `json:"content"`
+ Refusal string `json:"refusal"`
+}
+
+type ORChoice struct {
+ FinishReason string `json:"finish_reason"`
+ Index int `json:"index"`
+ Message ORMessage `json:"message"`
+}
+
+type ORUsage struct {
+ PromptTokens int `json:"prompt_tokens"`
+ CompletionTokens int `json:"completion_tokens"`
+ TotalTokens int `json:"total_tokens"`
+}
+type ORResponse struct {
+ Id string `json:"id"`
+ Provider string `json:"provider"`
+ Model string `json:"model"`
+ Object string `json:"object"`
+ Created int64 `json:"created"`
+ Choices []ORChoice `json:"choices"`
+ SystemFingerprint string `json:"system_fingerprint"`
+ Usage ORUsage `json:"usage"`
+}
+
type MemoryElement struct {
Role string `json:"role"`
Content string `json:"content"`