Merge branch 'main' into patch-1

README.md

@@ -29,8 +29,7 @@ curl https://ollama.ai/install.sh | sh
 
 ### Docker
 
-The official [Ollama Docker image `ollama/ollama`](https://hub.docker.com/r/ollama/ollama)
-is available on Docker Hub.
+The official [Ollama Docker image](https://hub.docker.com/r/ollama/ollama) `ollama/ollama` is available on Docker Hub.
 
 ## Quickstart
 
@@ -235,6 +234,7 @@ See the [API documentation](./docs/api.md) for all endpoints.
 - [LlamaIndex](https://gpt-index.readthedocs.io/en/stable/examples/llm/ollama.html)
 - [LiteLLM](https://github.com/BerriAI/litellm)
 - [OllamaSharp for .NET](https://github.com/awaescher/OllamaSharp)
+- [Ollama-rs for Rust](https://github.com/pepperoni21/ollama-rs)
 
 ### Plugins (Extensions)
 - [Raycast extension](https://github.com/MassimilianoPasquini97/raycast_ollama)
@@ -245,5 +245,3 @@ See the [API documentation](./docs/api.md) for all endpoints.
 - [Discord AI Bot](https://github.com/mekb-turtle/discord-ai-bot)
 - [Dumbar](https://github.com/JerrySievert/Dumbar)
 
-
-

api/client.go

@@ -72,7 +72,7 @@ func ClientFromEnvironment() (*Client, error) {
 		},
 	}
 
-	mockRequest, err := http.NewRequest("HEAD", client.base.String(), nil)
+	mockRequest, err := http.NewRequest(http.MethodHead, client.base.String(), nil)
 	if err != nil {
 		return nil, err
 	}

api/types.go

@@ -293,7 +293,7 @@ func DefaultOptions() Options {
 	return Options{
 		// options set on request to runner
 		NumPredict:       -1,
-		NumKeep:          -1,
+		NumKeep:          0,
 		Temperature:      0.8,
 		TopK:             40,
 		TopP:             0.9,

cmd/cmd.go

@@ -11,6 +11,7 @@ import (
 	"io"
 	"log"
 	"net"
+	"net/http"
 	"os"
 	"os/exec"
 	"os/signal"
@@ -98,19 +99,16 @@ func RunHandler(cmd *cobra.Command, args []string) error {
 		return err
 	}
 
-	models, err := client.List(context.Background())
-	if err != nil {
-		return err
-	}
-
-	canonicalModelPath := server.ParseModelPath(args[0])
-	for _, model := range models.Models {
-		if model.Name == canonicalModelPath.GetShortTagname() {
-			return RunGenerate(cmd, args)
+	name := args[0]
+	// check if the model exists on the server
+	_, err = client.Show(context.Background(), &api.ShowRequest{Name: name})
+	var statusError api.StatusError
+	switch {
+	case errors.As(err, &statusError) && statusError.StatusCode == http.StatusNotFound:
+		if err := PullHandler(cmd, args); err != nil {
+			return err
 		}
-	}
-
-	if err := PullHandler(cmd, args); err != nil {
+	case err != nil:
 		return err
 	}
 
@@ -731,21 +729,6 @@ func RunServer(cmd *cobra.Command, _ []string) error {
 		origins = strings.Split(o, ",")
 	}
 
-	if noprune := os.Getenv("OLLAMA_NOPRUNE"); noprune == "" {
-		if err := server.PruneLayers(); err != nil {
-			return err
-		}
-
-		manifestsPath, err := server.GetManifestPath()
-		if err != nil {
-			return err
-		}
-
-		if err := server.PruneDirectory(manifestsPath); err != nil {
-			return err
-		}
-	}
-
 	return server.Serve(ln, origins)
 }
 

docs/api.md

@@ -45,9 +45,11 @@ Advanced parameters (optional):
 - `system`: system prompt to (overrides what is defined in the `Modelfile`)
 - `template`: the full prompt or prompt template (overrides what is defined in the `Modelfile`)
 - `context`: the context parameter returned from a previous request to `/generate`, this can be used to keep a short conversational memory
-- `stream`: if `false` the response will be be returned as a single response object, rather than a stream of objects
+- `stream`: if `false` the response will be returned as a single response object, rather than a stream of objects
 
-### Request
+### Examples
+
+#### Request
 
 ```shell
 curl -X POST http://localhost:11434/api/generate -d '{
@@ -56,9 +58,9 @@ curl -X POST http://localhost:11434/api/generate -d '{
 }'
 ```
 
-### Response
+#### Response
 
-A stream of JSON objects:
+A stream of JSON objects is returned:
 
 ```json
 {
@@ -102,6 +104,38 @@ To calculate how fast the response is generated in tokens per second (token/s),
 }
 ```
 
+#### Request
+
+```shell
+curl -X POST http://localhost:11434/api/generate -d '{
+  "model": "llama2:7b",
+  "prompt": "Why is the sky blue?",
+  "stream": false
+}'
+```
+
+#### Response
+
+If `stream` is set to `false`, the response will be a single JSON object:
+
+```json
+{
+  "model": "llama2:7b",
+  "created_at": "2023-08-04T19:22:45.499127Z",
+  "response": "The sky is blue because it is the color of the sky.",
+  "context": [1, 2, 3],
+  "done": true,
+  "total_duration": 5589157167,
+  "load_duration": 3013701500,
+  "sample_count": 114,
+  "sample_duration": 81442000,
+  "prompt_eval_count": 46,
+  "prompt_eval_duration": 1160282000,
+  "eval_count": 13,
+  "eval_duration": 1325948000
+}
+```
+
 ## Create a Model
 
 ```shell
@@ -114,9 +148,11 @@ Create a model from a [`Modelfile`](./modelfile.md)
 
 - `name`: name of the model to create
 - `path`: path to the Modelfile
-- `stream`: (optional) if `false` the response will be be returned as a single response object, rather than a stream of objects
+- `stream`: (optional) if `false` the response will be returned as a single response object, rather than a stream of objects
+
+### Examples
 
-### Request
+#### Request
 
 ```shell
 curl -X POST http://localhost:11434/api/create -d '{
@@ -125,7 +161,7 @@ curl -X POST http://localhost:11434/api/create -d '{
 }'
 ```
 
-### Response
+#### Response
 
 A stream of JSON objects. When finished, `status` is `success`.
 
@@ -143,13 +179,17 @@ GET /api/tags
 
 List models that are available locally.
 
-### Request
+### Examples
+
+#### Request
 
 ```shell
 curl http://localhost:11434/api/tags
 ```
 
-### Response
+#### Response
+
+A single JSON object will be returned.
 
 ```json
 {
@@ -180,7 +220,9 @@ Show details about a model including modelfile, template, parameters, license, a
 
 - `name`: name of the model to show
 
-### Request
+### Examples
+
+#### Request
 
 ```shell
 curl http://localhost:11434/api/show -d '{
@@ -188,7 +230,7 @@ curl http://localhost:11434/api/show -d '{
 }'
 ```
 
-### Response
+#### Response
 
 ```json
 {
@@ -207,7 +249,9 @@ POST /api/copy
 
 Copy a model. Creates a model with another name from an existing model.
 
-### Request
+### Examples
+
+#### Request
 
 ```shell
 curl http://localhost:11434/api/copy -d '{
@@ -216,6 +260,10 @@ curl http://localhost:11434/api/copy -d '{
 }'
 ```
 
+#### Response
+
+The only response is a 200 OK if successful.
+
 ## Delete a Model
 
 ```shell
@@ -226,9 +274,11 @@ Delete a model and its data.
 
 ### Parameters
 
-- `model`: model name to delete
+- `name`: model name to delete
 
-### Request
+### Examples
+
+#### Request
 
 ```shell
 curl -X DELETE http://localhost:11434/api/delete -d '{
@@ -236,6 +286,10 @@ curl -X DELETE http://localhost:11434/api/delete -d '{
 }'
 ```
 
+#### Response
+
+If successful, the only response is a 200 OK.
+
 ## Pull a Model
 
 ```shell
@@ -248,9 +302,11 @@ Download a model from the ollama library. Cancelled pulls are resumed from where
 
 - `name`: name of the model to pull
 - `insecure`: (optional) allow insecure connections to the library. Only use this if you are pulling from your own library during development.
-- `stream`: (optional) if `false` the response will be be returned as a single response object, rather than a stream of objects
+- `stream`: (optional) if `false` the response will be returned as a single response object, rather than a stream of objects
+
+### Examples
 
-### Request
+#### Request
 
 ```shell
 curl -X POST http://localhost:11434/api/pull -d '{
@@ -258,13 +314,51 @@ curl -X POST http://localhost:11434/api/pull -d '{
 }'
 ```
 
-### Response
+#### Response
+
+If `stream` is not specified, or set to `true`, a stream of JSON objects is returned:
+
+The first object is the manifest:
+
+```json
+{
+  "status": "pulling manifest"
+}
+```
+
+Then there is a series of downloading responses. Until any of the download is completed, the `completed` key may not be included. The number of files to be downloaded depends on the number of layers specified in the manifest.
 
 ```json
 {
   "status": "downloading digestname",
   "digest": "digestname",
-  "total": 2142590208
+  "total": 2142590208,
+  "completed": 241970
+}
+```
+
+After all the files are downloaded, the final responses are:
+
+```json
+{
+    "status": "verifying sha256 digest"
+}
+{
+    "status": "writing manifest"
+}
+{
+    "status": "removing any unused layers"
+}
+{
+    "status": "success"
+}
+```
+
+if `stream` is set to false, then the response is a single JSON object:
+
+```json
+{
+  "status": "success"
 }
 ```
 
@@ -280,9 +374,11 @@ Upload a model to a model library. Requires registering for ollama.ai and adding
 
 - `name`: name of the model to push in the form of `<namespace>/<model>:<tag>`
 - `insecure`: (optional) allow insecure connections to the library. Only use this if you are pushing to your library during development.
-- `stream`: (optional) if `false` the response will be be returned as a single response object, rather than a stream of objects
+- `stream`: (optional) if `false` the response will be returned as a single response object, rather than a stream of objects
 
-### Request
+### Examples
+
+#### Request
 
 ```shell
 curl -X POST http://localhost:11434/api/push -d '{
@@ -290,9 +386,9 @@ curl -X POST http://localhost:11434/api/push -d '{
 }'
 ```
 
-### Response
+#### Response
 
-Streaming response that starts with:
+If `stream` is not specified, or set to `true`, a stream of JSON objects is returned:
 
 ```json
 { "status": "retrieving manifest" }
@@ -325,6 +421,12 @@ Finally, when the upload is complete:
 {"status":"success"}
 ```
 
+If `stream` is set to `false`, then the response is a single JSON object:
+
+```json
+{ "status": "success" }
+```
+
 ## Generate Embeddings
 
 ```shell
@@ -342,7 +444,9 @@ Advanced parameters:
 
 - `options`: additional model parameters listed in the documentation for the [Modelfile](./modelfile.md#valid-parameters-and-values) such as `temperature`
 
-### Request
+### Examples
+
+#### Request
 
 ```shell
 curl -X POST http://localhost:11434/api/embeddings -d '{
@@ -351,7 +455,7 @@ curl -X POST http://localhost:11434/api/embeddings -d '{
 }'
 ```
 
-### Response
+#### Response
 
 ```json
 {

docs/import.md

@@ -185,7 +185,7 @@ python convert.py <path to model directory>
 python convert-falcon-hf-to-gguf.py <path to model directory>
 
 # GPTNeoXForCausalLM
-python convert-falcon-hf-to-gguf.py <path to model directory>
+python convert-gptneox-hf-to-gguf.py <path to model directory>
 
 # GPTBigCodeForCausalLM
 python convert-starcoder-hf-to-gguf.py <path to model directory>

examples/langchain-python-rag-privategpt/constants.py

@@ -6,7 +6,6 @@ PERSIST_DIRECTORY = os.environ.get('PERSIST_DIRECTORY', 'db')
 
 # Define the Chroma settings
 CHROMA_SETTINGS = Settings(
-        chroma_db_impl='duckdb+parquet',
         persist_directory=PERSIST_DIRECTORY,
         anonymized_telemetry=False
 )

examples/langchain-python-rag-privategpt/ingest.py

@@ -150,7 +150,7 @@ def main():
         print("Creating new vectorstore")
         texts = process_documents()
         print(f"Creating embeddings. May take some minutes...")
-        db = Chroma.from_documents(texts, embeddings, persist_directory=persist_directory, client_settings=CHROMA_SETTINGS)
+        db = Chroma.from_documents(texts, embeddings, persist_directory=persist_directory)
     db.persist()
     db = None
 

examples/langchain-python-rag-privategpt/privateGPT.py

@@ -4,6 +4,7 @@ from langchain.embeddings import HuggingFaceEmbeddings
 from langchain.callbacks.streaming_stdout import StreamingStdOutCallbackHandler
 from langchain.vectorstores import Chroma
 from langchain.llms import Ollama
+import chromadb
 import os
 import argparse
 import time
@@ -22,7 +23,9 @@ def main():
     # Parse the command line arguments
     args = parse_arguments()
     embeddings = HuggingFaceEmbeddings(model_name=embeddings_model_name)
-    db = Chroma(persist_directory=persist_directory, embedding_function=embeddings, client_settings=CHROMA_SETTINGS)
+
+    db = Chroma(persist_directory=persist_directory, embedding_function=embeddings)
+
     retriever = db.as_retriever(search_kwargs={"k": target_source_chunks})
     # activate/deactivate the streaming StdOut callback for LLMs
     callbacks = [] if args.mute_stream else [StreamingStdOutCallbackHandler()]

go.mod

@@ -11,7 +11,6 @@ require (
 	github.com/olekukonko/tablewriter v0.0.5
 	github.com/spf13/cobra v1.7.0
 	golang.org/x/sync v0.3.0
-	gonum.org/v1/gonum v0.14.0
 )
 
 require github.com/rivo/uniseg v0.2.0 // indirect

go.sum

@@ -140,8 +140,6 @@ golang.org/x/text v0.13.0 h1:ablQoSUd0tRdKxZewP80B+BaqeKJuVhuRxj/dkrun3k=
 golang.org/x/text v0.13.0/go.mod h1:TvPlkZtksWOMsz7fbANvkp4WM8x/WCo/om8BMLbz+aE=
 golang.org/x/tools v0.0.0-20180917221912-90fa682c2a6e/go.mod h1:n7NCudcB/nEzxVGmLbDWY5pfWTLqBcC2KZ6jyYvM4mQ=
 golang.org/x/xerrors v0.0.0-20191204190536-9bdfabe68543/go.mod h1:I/5z698sn9Ka8TeJc9MKroUUfqBBauWjQqLJ2OPfmY0=
-gonum.org/v1/gonum v0.14.0 h1:2NiG67LD1tEH0D7kM+ps2V+fXmsAnpUeec7n8tcr4S0=
-gonum.org/v1/gonum v0.14.0/go.mod h1:AoWeoz0becf9QMWtE8iWXNXc27fK4fNeHNf/oMejGfU=
 google.golang.org/protobuf v1.26.0-rc.1/go.mod h1:jlhhOSvTdKEhbULTjvd4ARK9grFBp09yW+WbY/TyQbw=
 google.golang.org/protobuf v1.28.0/go.mod h1:HV8QOd/L58Z+nl8r43ehVNZIU/HEI6OcFqwMG9pJV4I=
 google.golang.org/protobuf v1.30.0 h1:kPPoIgf3TsEvrm0PFe15JQ+570QVxYzEvvHqChK+cng=

llm/llama.go

@@ -306,13 +306,19 @@ func newLlama(model string, adapters []string, runners []ModelRunner, numLayers
 	params := []string{
 		"--model", model,
 		"--ctx-size", fmt.Sprintf("%d", opts.NumCtx),
-		"--rope-freq-base", fmt.Sprintf("%f", opts.RopeFrequencyBase),
-		"--rope-freq-scale", fmt.Sprintf("%f", opts.RopeFrequencyScale),
 		"--batch-size", fmt.Sprintf("%d", opts.NumBatch),
 		"--n-gpu-layers", fmt.Sprintf("%d", numGPU),
 		"--embedding",
 	}
 
+	if opts.RopeFrequencyBase > 0 {
+		params = append(params, "--rope-freq-base", fmt.Sprintf("%f", opts.RopeFrequencyBase))
+	}
+
+	if opts.RopeFrequencyScale > 0 {
+		params = append(params, "--rope-freq-scale", fmt.Sprintf("%f", opts.RopeFrequencyScale))
+	}
+
 	if opts.NumGQA > 0 {
 		params = append(params, "--gqa", fmt.Sprintf("%d", opts.NumGQA))
 	}
@@ -360,7 +366,15 @@ func newLlama(model string, adapters []string, runners []ModelRunner, numLayers
 			runner.Path,
 			append(params, "--port", strconv.Itoa(port))...,
 		)
-		cmd.Env = append(os.Environ(), fmt.Sprintf("LD_LIBRARY_PATH=%s", filepath.Dir(runner.Path)))
+
+		var libraryPaths []string
+		if libraryPath, ok := os.LookupEnv("LD_LIBRARY_PATH"); ok {
+			libraryPaths = append(libraryPaths, libraryPath)
+		}
+
+		libraryPaths = append(libraryPaths, filepath.Dir(runner.Path))
+
+		cmd.Env = append(os.Environ(), fmt.Sprintf("LD_LIBRARY_PATH=%s", strings.Join(libraryPaths, ":")))
 		cmd.Stdout = os.Stderr
 		statusWriter := NewStatusWriter()
 		cmd.Stderr = statusWriter

llm/llm.go

@@ -85,7 +85,10 @@ func New(workDir, model string, adapters []string, opts api.Options) (LLM, error
 
 	switch ggml.Name() {
 	case "gguf":
-		opts.NumGQA = 0 // TODO: remove this when llama.cpp runners differ enough to need separate newLlama functions
+		// TODO: gguf will load these options automatically from the model binary
+		opts.NumGQA = 0
+		opts.RopeFrequencyBase = 0.0
+		opts.RopeFrequencyScale = 0.0
 		return newLlama(model, adapters, chooseRunners(workDir, "gguf"), ggml.NumLayers(), opts)
 	case "ggml", "ggmf", "ggjt", "ggla":
 		return newLlama(model, adapters, chooseRunners(workDir, "ggml"), ggml.NumLayers(), opts)

readline/buffer.go

@@ -2,6 +2,7 @@ package readline
 
 import (
 	"fmt"
+	"os"
 
 	"github.com/emirpasic/gods/lists/arraylist"
 	"golang.org/x/term"
@@ -17,7 +18,8 @@ type Buffer struct {
 }
 
 func NewBuffer(prompt *Prompt) (*Buffer, error) {
-	width, height, err := term.GetSize(0)
+	fd := int(os.Stdout.Fd())
+	width, height, err := term.GetSize(fd)
 	if err != nil {
 		fmt.Println("Error getting size:", err)
 		return nil, err

readline/readline.go

@@ -51,11 +51,12 @@ func (i *Instance) Readline() (string, error) {
 	}
 	fmt.Print(prompt)
 
-	termios, err := SetRawMode(syscall.Stdin)
+	fd := int(syscall.Stdin)
+	termios, err := SetRawMode(fd)
 	if err != nil {
 		return "", err
 	}
-	defer UnsetRawMode(syscall.Stdin, termios)
+	defer UnsetRawMode(fd, termios)
 
 	buf, _ := NewBuffer(i.Prompt)
 

readline/term_bsd.go

 //go:build darwin || freebsd || netbsd || openbsd
+
 package readline
 
 import (

readline/term_linux.go

 //go:build linux || solaris
+
 package readline
 
 import (

readline/term_windows.go

+package readline
+
+import (
+	"syscall"
+	"unsafe"
+)
+
+const (
+	enableLineInput       = 2
+	enableWindowInput     = 8
+	enableMouseInput      = 16
+	enableInsertMode      = 32
+	enableQuickEditMode   = 64
+	enableExtendedFlags   = 128
+	enableProcessedOutput = 1
+	enableWrapAtEolOutput = 2
+	enableAutoPosition    = 256 // Cursor position is not affected by writing data to the console.
+	enableEchoInput       = 4   // Characters are written to the console as they're read.
+	enableProcessedInput  = 1   // Enables input processing (like recognizing Ctrl+C).
+)
+
+var kernel32 = syscall.NewLazyDLL("kernel32.dll")
+
+var (
+	procGetConsoleMode = kernel32.NewProc("GetConsoleMode")
+	procSetConsoleMode = kernel32.NewProc("SetConsoleMode")
+)
+
+type State struct {
+	mode uint32
+}
+
+// IsTerminal checks if the given file descriptor is associated with a terminal
+func IsTerminal(fd int) bool {
+	var st uint32
+	r, _, e := syscall.SyscallN(procGetConsoleMode.Addr(), uintptr(fd), uintptr(unsafe.Pointer(&st)), 0)
+	// if the call succeeds and doesn't produce an error, it's a terminal
+	return r != 0 && e == 0
+}
+
+func SetRawMode(fd int) (*State, error) {
+	var st uint32
+	// retrieve the current mode of the terminal
+	_, _, e := syscall.SyscallN(procGetConsoleMode.Addr(), uintptr(fd), uintptr(unsafe.Pointer(&st)), 0)
+	if e != 0 {
+		return nil, error(e)
+	}
+	// modify the mode to set it to raw
+	raw := st &^ (enableEchoInput | enableProcessedInput | enableLineInput | enableProcessedOutput)
+	// apply the new mode to the terminal
+	_, _, e = syscall.SyscallN(procSetConsoleMode.Addr(), uintptr(fd), uintptr(raw), 0)
+	if e != 0 {
+		return nil, error(e)
+	}
+	// return the original state so that it can be restored later
+	return &State{st}, nil
+}
+
+func UnsetRawMode(fd int, state *State) error {
+	_, _, err := syscall.SyscallN(procSetConsoleMode.Addr(), uintptr(fd), uintptr(state.mode), 0)
+	return err
+}

scripts/install.sh

@@ -63,7 +63,10 @@ status "Installing ollama to $BINDIR..."
 $SUDO install -o0 -g0 -m755 -d $BINDIR
 $SUDO install -o0 -g0 -m755 $TEMP_DIR/ollama $BINDIR/ollama
 
-install_success() { status 'Install complete. Run "ollama" from the command line.'; }
+install_success() { 
+    status 'The Ollama API is now available at 0.0.0.0:11434.'
+    status 'Install complete. Run "ollama" from the command line.'
+}
 trap install_success EXIT
 
 # Everything from this point onwards is optional.
@@ -130,6 +133,7 @@ if check_gpu nvidia-smi; then
 fi
 
 if ! check_gpu lspci && ! check_gpu lshw; then
+    install_success
     warning "No NVIDIA GPU detected. Ollama will run in CPU-only mode."
     exit 0
 fi