main.nim improved and used as default app.

Author: Judd

README.md

@@ -42,7 +42,7 @@ pure C++ implementation based on [@ggerganov](https://github.com/ggerganov)'s [g
 
 ## Quick Start
 
-As simple as `python chatllm.py -i -m :model_id`. [Check it out](./docs/quick_start.md).
+As simple as `main -i -m :model_id`. [Check it out](./docs/quick_start.md).
 
 ## Usage
 

README_ja.md

@@ -10,19 +10,6 @@
 
 | [サポートされているモデル](./docs/models.md) | [量子化モデルのダウンロード](./docs/quick_start.md#download-quantized-models) |
 
-**新着情報:**
-
-* 2024-11-01: Granite, [生成ステアリング](./docs/fun.md#generation-steering)
-* 2024-09-29: LlaMA 3.2
-* 2024-09-22: Qwen 2.5
-* 2024-09-13: OLMoE
-* 2024-09-11: MiniCPM3
-* 2024-07-14: [ggml 更新](https://github.com/ggerganov/ggml/tree/3e7e5e26f90fecf4f7c2808df7d94454630b219c)
-* 2024-06-15: [ツール呼び出し](./docs/tool_calling.md)
-* 2024-05-29: [ggml](https://github.com/ggerganov/ggml) はサブモジュールではなくフォークされました
-* 2024-05-14: [OpenAI API](./docs/binding.md#openai-compatible-api), CodeGemma Base & Instruct 対応
-* 2024-05-08: [レイヤーシャッフル](./docs/fun.md#layer-shuffling)
-
 ## 特徴
 
 * [x] int4/int8 量子化、最適化された KV キャッシュ、並列計算によるメモリ効率の高い CPU 推論の加速

README_zh.md

@@ -26,7 +26,7 @@
 
 ## 快速开始
 
-只需要简单一行 `python chatllm.py -i -m :model_id`. 查看 [详情](./docs/quick_start.md).
+只需要简单一行 `main -i -m :model_id`. 查看 [详情](./docs/quick_start.md).
 
 ## 使用方法
 

bindings/main.nim

@@ -147,7 +147,17 @@ var storage_dir: string = "../quantized"
 var ht = highlighter(line_acc: "", lang: langNone)
 let chat = chatllm_create()
 
+# related front end parameters are ignored by `libchatllm`
+var prompt: string = "hello"
+var interactive: bool = false
+
 for i in 1 .. paramCount():
+  if paramStr(i) in ["-i", "--interactive"]:
+    interactive = true
+
+  if (i > 1) and (paramStr(i - 1) in ["-p", "--prompt"]):
+    prompt = paramStr(i)
+
   if (i > 1) and (paramStr(i - 1) in candidates) and paramStr(i).startsWith(":"):
     var m = paramStr(i)
     m = m[1..<len(m)]
@@ -160,13 +170,18 @@ if r != 0:
   echo ">>> chatllm_start error: ", r
   quit(r)
 
-while true:
-  stdout.write("You  > ")
-  let input = stdin.readLine()
-  if input.isEmptyOrWhitespace(): continue
-
-  stdout.write("A.I. > ")
-  let r = chatllm_user_input(chat, input.cstring)
-  if r != 0:
-    echo ">>> chatllm_user_input error: ", r
-    break
+if interactive:
+  while true:
+    stdout.write("You  > ")
+    let input = stdin.readLine()
+    if input.isEmptyOrWhitespace(): continue
+
+    stdout.write("A.I. > ")
+    let r = chatllm_user_input(chat, input.cstring)
+    if r != 0:
+      echo ">>> chatllm_user_input error: ", r
+      break
+else:
+  discard chatllm_user_input(chat, prompt.cstring)
+
+chatllm_show_statistics(chat)

docs/quick_start.md

@@ -3,8 +3,8 @@
 For Windows users, the easies way is to download a release, extract it, and start chatting:
 
 ```
-python chatllm.py -i -m :qwen2:0.5b
-downloading qwen2:0.5b
+main -i -m :qwen2.5:0.5b
+Downloading qwen2:0.5b
  |████████████████████████████████████████████████████████████| 100.0%
     ________          __  __    __    __  ___ (通义千问)
    / ____/ /_  ____ _/ /_/ /   / /   /  |/  /_________  ____