;============================================================ ; iron_ai_vision.hsp — マルチモーダル (画像入力) 拡張 ; ; GPT-4o / Claude / Llama 3.2 Vision 等のビジョン対応モデルに ; 画像を送って応答を得る薄いラッパ。 ; ; 依存: iron_ai.hsp + iron_hash.hsp ; ; API: ; iron_ai_vision_file "image.png", "この画像を説明して", reply ; ローカル画像ファイルを Base64 エンコードして送信。 ; stat: 200=成功 ; ; iron_ai_vision_base64 base64_str, "prompt", reply ; 既に Base64 化済みの画像データで問い合わせ。 ; ; iron_ai_vision_url "https://...", "prompt", reply ; 画像 URL を指定して問い合わせ。 ; ; 使用例: ; #include "iron_ai_vision.hsp" ; ; iron_ai_set_endpoint "https://api.openai.com/v1" ; iron_ai_set_key "sk-..." ; iron_ai_set_model "gpt-4o" ; iron_ai_vision_file "photo.jpg", "何が写っていますか？", reply ; if stat = 200 : mes reply ;============================================================ #ifndef __iron_ai_vision_hsp__ #define __iron_ai_vision_hsp__ #include "iron_ai.hsp" #include "iron_hash.hsp" #module iron_ai_vision ;------------------------------------------------------------ ; 内部: JSON 文字列エスケープ ;------------------------------------------------------------ #deffunc _aiv_json_escape str src, var dst, local _i, local _c, local _len, local _src _src = src _len = strlen(_src) sdim dst, _len * 6 + 16 repeat _len _c = peek(_src, cnt) if _c = '\\' { poke dst, strlen(dst), '\\' poke dst, strlen(dst), '\\' } else : if _c = '"' { poke dst, strlen(dst), '\\' poke dst, strlen(dst), '"' } else : if _c = 0x0a { poke dst, strlen(dst), '\\' poke dst, strlen(dst), 'n' } else : if _c = 0x0d { poke dst, strlen(dst), '\\' poke dst, strlen(dst), 'r' } else : if _c = 0x09 { poke dst, strlen(dst), '\\' poke dst, strlen(dst), 't' } else : if _c < 0x20 { ; skip } else { poke dst, strlen(dst), _c } loop return ;------------------------------------------------------------ ; 内部: 拡張子から MIME タイプを推定 ;------------------------------------------------------------ #defcfunc _aiv_mime str path, local _ext _ext = getpath(path, 18) if _ext = ".png" : return "image/png" if _ext = ".jpg" : return "image/jpeg" if _ext = ".jpeg" : return "image/jpeg" if _ext = ".gif" : return "image/gif" if _ext = ".webp" : return "image/webp" if _ext = ".bmp" : return "image/bmp" return "image/png" ;------------------------------------------------------------ ; 内部: Base64 画像 + テキストプロンプトで vision API 呼び出し ; base64_data: Base64 エンコード済み画像データ ; mime: MIME タイプ ("image/png" 等) ; prompt: テキストプロンプト ; reply: 応答格納先 ;------------------------------------------------------------ #deffunc _aiv_chat_base64 str base64_data, str mime, str prompt, var reply, local _esc_prompt, local _sys_esc, local _body, local _msgs, local _full_url, local _resp, local _hid, local _hdr_buf, local _content_arr sdim reply, 4096 reply = "" _aiv_json_escape prompt, _esc_prompt ; content 配列: [image_url, text] sdim _content_arr, strlen(base64_data) + strlen(_esc_prompt) + 1024 _content_arr = "[{\"type\":\"image_url\",\"image_url\":{\"url\":\"data:" + mime + ";base64," + base64_data + "\"}}," _content_arr += "{\"type\":\"text\",\"text\":\"" + _esc_prompt + "\"}]" ; messages 配列 sdim _msgs, strlen(_content_arr) + strlen(_ai_system@iron_ai) * 2 + strlen(_ai_history_json@iron_ai) + 2048 _msgs = "" if strlen(_ai_system@iron_ai) > 0 { _aiv_json_escape _ai_system@iron_ai, _sys_esc _msgs = "{\"role\":\"system\",\"content\":\"" + _sys_esc + "\"}," } if strlen(_ai_history_json@iron_ai) > 0 { _msgs = _msgs + _ai_history_json@iron_ai + "," } _msgs = _msgs + "{\"role\":\"user\",\"content\":" + _content_arr + "}" ; body sdim _body, strlen(_msgs) + 1024 _body = "{\"model\":\"" + _ai_model@iron_ai + "\"," _body += "\"messages\":[" + _msgs + "]" if _ai_max_tokens@iron_ai > 0 { _body += ",\"max_tokens\":" + _ai_max_tokens@iron_ai } _body += "}" ; Authorization sdim _hdr_buf, 2048 if strlen(_ai_key@iron_ai) > 0 { _hdr_buf = "Authorization: Bearer " + _ai_key@iron_ai + "\r\n" } else { _hdr_buf = "" } http_set_header _hdr_buf sdim _full_url, 1024 _full_url = _ai_endpoint@iron_ai + "/chat/completions" sdim _resp, 1 http_post _full_url, _body, _resp, "application/json" if stat ! 200 : return stat _hid = json_load(_resp) if _hid < 0 : return -1 reply = json_str(_hid, "choices[0].message.content") json_release _hid ; 履歴に追加 iron_ai_history_add "user", prompt iron_ai_history_add "assistant", reply return 200 ;------------------------------------------------------------ ; iron_ai_vision_file "image.png", "prompt", reply ; ファイルを読み込み Base64 エンコードして送信。 ; stat: 200=成功 ;------------------------------------------------------------ #deffunc iron_ai_vision_file str image_path, str prompt, var reply, local _file_buf, local _file_len, local _b64, local _mime sdim reply, 4096 reply = "" ; ファイル存在確認 exist image_path if strsize < 0 : return -1 _file_len = strsize ; ファイル読み込み sdim _file_buf, _file_len + 16 bload image_path, _file_buf, _file_len ; Base64 エンコード (iron_hash の base64_encode_buf) base64_encode_buf _file_buf, _file_len _b64 = refstr ; MIME 推定 _mime = _aiv_mime(image_path) ; API 呼び出し _aiv_chat_base64 _b64, _mime, prompt, reply return stat ;------------------------------------------------------------ ; iron_ai_vision_base64 base64_str, "prompt", reply ; 既に Base64 化済みの画像で問い合わせ。 ; MIME は image/png として送信。 ;------------------------------------------------------------ #deffunc iron_ai_vision_base64 str base64_str, str prompt, var reply sdim reply, 4096 _aiv_chat_base64 base64_str, "image/png", prompt, reply return stat ;------------------------------------------------------------ ; iron_ai_vision_url "https://...", "prompt", reply ; 画像 URL を直接指定して問い合わせ。 ;------------------------------------------------------------ #deffunc iron_ai_vision_url str image_url, str prompt, var reply, local _esc_prompt, local _sys_esc, local _body, local _msgs, local _full_url, local _resp, local _hid, local _hdr_buf, local _content_arr, local _esc_url sdim reply, 4096 reply = "" _aiv_json_escape prompt, _esc_prompt _aiv_json_escape image_url, _esc_url ; content 配列 sdim _content_arr, strlen(_esc_url) + strlen(_esc_prompt) + 512 _content_arr = "[{\"type\":\"image_url\",\"image_url\":{\"url\":\"" + _esc_url + "\"}}," _content_arr += "{\"type\":\"text\",\"text\":\"" + _esc_prompt + "\"}]" ; messages sdim _msgs, strlen(_content_arr) + strlen(_ai_system@iron_ai) * 2 + strlen(_ai_history_json@iron_ai) + 2048 _msgs = "" if strlen(_ai_system@iron_ai) > 0 { _aiv_json_escape _ai_system@iron_ai, _sys_esc _msgs = "{\"role\":\"system\",\"content\":\"" + _sys_esc + "\"}," } if strlen(_ai_history_json@iron_ai) > 0 { _msgs = _msgs + _ai_history_json@iron_ai + "," } _msgs = _msgs + "{\"role\":\"user\",\"content\":" + _content_arr + "}" ; body sdim _body, strlen(_msgs) + 1024 _body = "{\"model\":\"" + _ai_model@iron_ai + "\"," _body += "\"messages\":[" + _msgs + "]" if _ai_max_tokens@iron_ai > 0 { _body += ",\"max_tokens\":" + _ai_max_tokens@iron_ai } _body += "}" ; Authorization sdim _hdr_buf, 2048 if strlen(_ai_key@iron_ai) > 0 { _hdr_buf = "Authorization: Bearer " + _ai_key@iron_ai + "\r\n" } else { _hdr_buf = "" } http_set_header _hdr_buf sdim _full_url, 1024 _full_url = _ai_endpoint@iron_ai + "/chat/completions" sdim _resp, 1 http_post _full_url, _body, _resp, "application/json" if stat ! 200 : return stat _hid = json_load(_resp) if _hid < 0 : return -1 reply = json_str(_hid, "choices[0].message.content") json_release _hid iron_ai_history_add "user", prompt iron_ai_history_add "assistant", reply return 200 #global #endif