| 1 | package responses |
| 2 | |
| 3 | import ( |
| 4 | "bytes" |
| 5 | "encoding/json" |
| 6 | "strings" |
| 7 | "testing" |
| 8 | |
| 9 | "reasonix/internal/provider" |
| 10 | ) |
| 11 | |
| 12 | func TestOfficialDeepSeekVisionSKUEmbedsUserImages(t *testing.T) { |
| 13 | c := New(Config{ |
| 14 | Name: "deepseek", BaseURL: "https://api.deepseek.com", Model: "deepseek-v4-flash-vision-exp", |
| 15 | }).(*client) |
| 16 | if !c.vision { |
| 17 | t.Fatal("pinned official DeepSeek vision SKU must enable user image serialization") |
| 18 | } |
| 19 | body, _, _ := c.buildRequestBody(provider.Request{Messages: []provider.Message{{ |
| 20 | Role: provider.RoleUser, Content: "what is this", |
| 21 | Images: []string{"data:image/png;base64,AAAA"}, |
| 22 | }}}) |
| 23 | items := body["input"].([]map[string]any) |
| 24 | parts, ok := items[0]["content"].([]map[string]string) |
| 25 | if !ok || len(parts) != 2 || parts[0]["type"] != "input_text" || parts[1]["type"] != "input_image" { |
| 26 | t.Fatalf("vision SKU content = %#v, want input_text + input_image", items[0]["content"]) |
| 27 | } |
| 28 | textOnly, _, _ := c.buildRequestBody(provider.Request{Messages: []provider.Message{{ |
| 29 | Role: provider.RoleUser, Content: "hello", |
| 30 | }}}) |
| 31 | if got, ok := textOnly["input"].([]map[string]any)[0]["content"].(string); !ok || got != "hello" { |
| 32 | t.Fatalf("vision SKU text-only content = %#v, want a string", textOnly["input"].([]map[string]any)[0]["content"]) |
| 33 | } |
| 34 | } |
| 35 | |
| 36 | func TestOfficialRequestURLImageHardLimit(t *testing.T) { |
| 37 | c := New(Config{BaseURL: "https://relay.test", RequestURL: "https://api.deepseek.com/responses", Model: "deepseek-v4-pro", Extra: map[string]any{"vision": true}, ModelInfo: &provider.ModelInfo{InputModalities: []provider.ModelModality{provider.ModalityText, provider.ModalityImage}}}).(*client) |
| 38 | req, _, _ := c.buildRequestBody(provider.Request{Messages: []provider.Message{{Role: provider.RoleUser, Content: "describe", Images: []string{"data:image/png;base64,AAAA"}}}}) |
| 39 | body, err := json.Marshal(req) |
| 40 | if err != nil || strings.Contains(string(body), "AAAA") { |
| 41 | t.Fatalf("official request URL leaked image: %s %v", body, err) |
| 42 | } |
| 43 | } |
| 44 | |
| 45 | func TestOfficialVisionExplicitOffRespectsResolvedMetadata(t *testing.T) { |
| 46 | c := New(Config{BaseURL: "https://api.deepseek.com", Model: "deepseek-v4-flash-vision-exp", Extra: map[string]any{"vision": true}, ModelInfo: &provider.ModelInfo{InputModalities: []provider.ModelModality{provider.ModalityText}}}).(*client) |
| 47 | req, _, _ := c.buildRequestBody(provider.Request{Messages: []provider.Message{{Role: provider.RoleUser, Content: "describe", Images: []string{"data:image/png;base64,AAAA"}}}}) |
| 48 | body, err := json.Marshal(req) |
| 49 | if err != nil || strings.Contains(string(body), "AAAA") { |
| 50 | t.Fatalf("explicit off leaked image: %s %v", body, err) |
| 51 | } |
| 52 | if c.ModelInfo().SupportsInput(provider.ModalityImage) { |
| 53 | t.Fatal("metadata disagrees with serializer") |
| 54 | } |
| 55 | } |
| 56 | |
| 57 | func TestModelInfoEnablesImageWireSerialization(t *testing.T) { |
| 58 | c := New(Config{ |
| 59 | Name: "catalog", BaseURL: "https://example.test/v1", Model: "vision", |
| 60 | ModelInfo: &provider.ModelInfo{ID: "vision", InputModalities: []provider.ModelModality{provider.ModalityText, provider.ModalityImage}}, |
| 61 | }).(*client) |
| 62 | if !c.vision { |
| 63 | t.Fatal("model metadata should enable image wire serialization") |
| 64 | } |
| 65 | body, _, _ := c.buildRequestBody(provider.Request{Messages: []provider.Message{{Role: provider.RoleUser, Content: "describe", Images: []string{"data:image/png;base64,AAAA"}}}}) |
| 66 | items := body["input"].([]map[string]any) |
| 67 | parts := items[0]["content"].([]map[string]string) |
| 68 | if len(parts) != 2 || parts[1]["type"] != "input_image" { |
| 69 | t.Fatalf("content = %#v, want text + input_image parts", parts) |
| 70 | } |
| 71 | } |
| 72 | |
| 73 | func TestOfficialDeepSeekVisionSKUEmbedsURLAndFileID(t *testing.T) { |
| 74 | c := New(Config{ |
| 75 | Name: "deepseek", BaseURL: "https://api.deepseek.com", Model: "deepseek-v4-flash-vision-exp", |
| 76 | }).(*client) |
| 77 | body, _, _ := c.buildRequestBody(provider.Request{Messages: []provider.Message{{ |
| 78 | Role: provider.RoleUser, |
| 79 | Content: "what is this", |
| 80 | Images: []string{ |
| 81 | "https://cdn.example.com/cat.png", |
| 82 | "file-api-0a1b2c3d4e5f6071", |
| 83 | }, |
| 84 | }}}) |
| 85 | items := body["input"].([]map[string]any) |
| 86 | parts, ok := items[0]["content"].([]map[string]string) |
| 87 | if !ok || len(parts) != 3 { |
| 88 | t.Fatalf("content = %#v", items[0]["content"]) |
| 89 | } |
| 90 | if parts[1]["type"] != "input_image" || parts[1]["image_url"] != "https://cdn.example.com/cat.png" { |
| 91 | t.Fatalf("url part = %#v", parts[1]) |
| 92 | } |
| 93 | if parts[2]["type"] != "input_image" || parts[2]["file_id"] != "file-api-0a1b2c3d4e5f6071" { |
| 94 | t.Fatalf("file part = %#v", parts[2]) |
| 95 | } |
| 96 | } |
| 97 | |
| 98 | func TestStatefulContinuationDoesNotDropUserImages(t *testing.T) { |
| 99 | c := New(Config{ |
| 100 | Name: "deepseek", BaseURL: "https://api.deepseek.com", Model: "deepseek-v4-flash-vision-exp", Mode: "stateful", |
| 101 | }).(*client) |
| 102 | prefix := []provider.Message{ |
| 103 | {Role: provider.RoleSystem, Content: "You inspect images."}, |
| 104 | {Role: provider.RoleUser, Content: "first"}, |
| 105 | {Role: provider.RoleAssistant, Content: "answer"}, |
| 106 | } |
| 107 | c.lastResponseID = "resp_1" |
| 108 | c.expectedPrefixDigest = c.conversationDigest(prefix) |
| 109 | body, usedPrevious, _ := c.buildRequestBody(provider.Request{Messages: append(prefix, provider.Message{ |
| 110 | Role: provider.RoleUser, Content: "second", Images: []string{"data:image/png;base64,AAAA"}, |
| 111 | })}) |
| 112 | if usedPrevious { |
| 113 | t.Fatal("stateful continuation must not use text-only previous_response_id fast path for image turns") |
| 114 | } |
| 115 | items, ok := body["input"].([]map[string]any) |
| 116 | if !ok { |
| 117 | t.Fatalf("input = %#v, want multimodal item list", body["input"]) |
| 118 | } |
| 119 | found := false |
| 120 | for _, item := range items { |
| 121 | parts, ok := item["content"].([]map[string]string) |
| 122 | if !ok { |
| 123 | continue |
| 124 | } |
| 125 | for _, part := range parts { |
| 126 | if part["type"] == "input_image" && part["image_url"] == "data:image/png;base64,AAAA" { |
| 127 | found = true |
| 128 | } |
| 129 | } |
| 130 | } |
| 131 | if !found { |
| 132 | t.Fatalf("stateful continuation input = %#v, missing user image", items) |
| 133 | } |
| 134 | } |
| 135 | |
| 136 | func TestOfficialDeepSeekVisionSKUEmbedsToolImages(t *testing.T) { |
| 137 | c := New(Config{ |
| 138 | Name: "deepseek", BaseURL: "https://api.deepseek.com", Model: "deepseek-v4-flash-vision-exp", |
| 139 | Extra: map[string]any{"vision": true}, |
| 140 | }).(*client) |
| 141 | plain := []provider.Message{ |
| 142 | {Role: provider.RoleUser, Content: "inspect"}, |
| 143 | {Role: provider.RoleAssistant, ToolCalls: []provider.ToolCall{{ID: "c1", Name: "shot", Arguments: "{}"}}}, |
| 144 | {Role: provider.RoleTool, ToolCallID: "c1", Name: "shot", Content: "vision result"}, |
| 145 | } |
| 146 | withToolImage := append([]provider.Message(nil), plain...) |
| 147 | withToolImage[2].Images = []string{"data:image/png;base64,VE9PTA=="} |
| 148 | plainBody, err := json.Marshal(mustRequest(t, c, plain)) |
| 149 | if err != nil { |
| 150 | t.Fatalf("marshal plain: %v", err) |
| 151 | } |
| 152 | imageBody, err := json.Marshal(mustRequest(t, c, withToolImage)) |
| 153 | if err != nil { |
| 154 | t.Fatalf("marshal image: %v", err) |
| 155 | } |
| 156 | if !strings.Contains(string(imageBody), "VE9PTA") { |
| 157 | t.Fatalf("official DeepSeek vision SKU omitted tool image payload: %s", imageBody) |
| 158 | } |
| 159 | if bytes.Equal(imageBody, plainBody) { |
| 160 | t.Fatalf("tool images changed official DeepSeek vision SKU bytes:\nplain: %s\nimage: %s", plainBody, imageBody) |
| 161 | } |
| 162 | } |
| 163 | |
| 164 | func mustRequest(t *testing.T, c *client, messages []provider.Message) map[string]any { |
| 165 | t.Helper() |
| 166 | body, _, _ := c.buildRequestBody(provider.Request{Messages: messages}) |
| 167 | return body |
| 168 | } |
| 169 | |
| 170 | func TestOfficialDeepSeekResponsesImageMetadataMatchesTextOnlyWireBytes(t *testing.T) { |
| 171 | c := New(Config{ |
| 172 | Name: "deepseek", BaseURL: "https://api.deepseek.com", Model: "deepseek-v4-pro", |
| 173 | Extra: map[string]any{"vision": true}, |
| 174 | }).(*client) |
| 175 | plain := []provider.Message{ |
| 176 | {Role: provider.RoleUser, Content: "inspect"}, |
| 177 | {Role: provider.RoleAssistant, ToolCalls: []provider.ToolCall{{ID: "c1", Name: "shot", Arguments: "{}"}}}, |
| 178 | {Role: provider.RoleTool, ToolCallID: "c1", Name: "shot", Content: "vision result"}, |
| 179 | } |
| 180 | withImages := append([]provider.Message(nil), plain...) |
| 181 | withImages[0].Images = []string{"data:image/png;base64," + strings.Repeat("QUFB", 20_000)} |
| 182 | withImages[2].Images = []string{"data:image/png;base64,VE9PTA=="} |
| 183 | |
| 184 | plainRequest, _, _ := c.buildRequestBody(provider.Request{Messages: plain}) |
| 185 | imageRequest, _, _ := c.buildRequestBody(provider.Request{Messages: withImages}) |
| 186 | plainBody, err := json.Marshal(plainRequest) |
| 187 | if err != nil { |
| 188 | t.Fatalf("marshal plain request: %v", err) |
| 189 | } |
| 190 | imageBody, err := json.Marshal(imageRequest) |
| 191 | if err != nil { |
| 192 | t.Fatalf("marshal image request: %v", err) |
| 193 | } |
| 194 | if !bytes.Equal(imageBody, plainBody) { |
| 195 | t.Fatalf("official DeepSeek Responses image metadata changed provider-visible bytes:\nplain: %s\nimage: %s", plainBody, imageBody) |
| 196 | } |
| 197 | } |
| 198 | |
| 199 | func TestToolImagesFollowCompleteToolRun(t *testing.T) { |
| 200 | for _, trailing := range []bool{false, true} { |
| 201 | for _, vision := range []bool{false, true} { |
| 202 | messages := []provider.Message{ |
| 203 | {Role: provider.RoleAssistant, ToolCalls: []provider.ToolCall{{ID: "a", Name: "view_image", Arguments: "{}"}, {ID: "b", Name: "other", Arguments: "{}"}}}, |
| 204 | {Role: provider.RoleTool, ToolCallID: "a", Content: "picture", Images: []string{"data:image/png;base64,AAAA", "invalid"}}, |
| 205 | {Role: provider.RoleTool, ToolCallID: "b", Content: "done"}, |
| 206 | } |
| 207 | if trailing { |
| 208 | messages = append(messages, provider.Message{Role: provider.RoleUser, Content: "next"}) |
| 209 | } |
| 210 | items := messagesToInput(messages, vision, false, false) |
| 211 | want := 4 |
| 212 | if vision { |
| 213 | want++ |
| 214 | } |
| 215 | if trailing { |
| 216 | want++ |
| 217 | } |
| 218 | if len(items) != want { |
| 219 | t.Fatalf("vision=%v trailing=%v: %#v", vision, trailing, items) |
| 220 | } |
| 221 | if items[2]["type"] != "function_call_output" || items[3]["call_id"] != "b" { |
| 222 | t.Fatalf("tool ordering: %#v", items) |
| 223 | } |
| 224 | if vision { |
| 225 | parts := items[4]["content"].([]map[string]string) |
| 226 | if items[4]["role"] != "user" || len(parts) != 2 || parts[1]["image_url"] != "data:image/png;base64,AAAA" { |
| 227 | t.Fatalf("image injection: %#v", items[4]) |
| 228 | } |
| 229 | } else { |
| 230 | encoded, _ := json.Marshal(items) |
| 231 | if bytes.Contains(encoded, []byte("AAAA")) { |
| 232 | t.Fatalf("image leaked to text model") |
| 233 | } |
| 234 | } |
| 235 | if trailing && items[len(items)-1]["content"] != "next" { |
| 236 | t.Fatal("displaced user message") |
| 237 | } |
| 238 | } |
| 239 | } |
| 240 | } |
| 241 |