| 1 | //! Deterministic JSON argument repair for malformed tool-call inputs. |
| 2 | //! |
| 3 | //! DeepSeek streams `tool_calls.function.arguments` as deltas. Two failure |
| 4 | //! shapes are common: (a) SSE chunk boundary cuts inside a JSON string and |
| 5 | //! reassembly leaves a trailing comma or unclosed brace; (b) some local |
| 6 | //! backends emit literal control characters inside JSON string values. |
| 7 | //! |
| 8 | //! The repair ladder runs five stages before reporting unrecoverable input: |
| 9 | //! |
| 10 | //! 1. Strict parse — done if it parses. |
| 11 | //! 2. Strip literal control chars inside string values. |
| 12 | //! 3. Strip trailing commas before `}` or `]`. |
| 13 | //! 4. Balance braces/brackets (append closers). |
| 14 | //! 5. Strip excess closers if delta is negative. |
| 15 | //! |
| 16 | //! Stages 1-3 never change structure: they parse as-is, or normalize text |
| 17 | //! that was already structurally complete. Stages 4-5 do — they synthesize |
| 18 | //! or discard closers to force a parse. A value that only parsed because of |
| 19 | //! stage 4 or 5 came from argument text that was *incomplete*, and the usual |
| 20 | //! cause is a provider cutting the stream at its output limit mid-argument. |
| 21 | //! `Repaired::structure_synthesized` reports that, because such a value must |
| 22 | //! never be dispatched as if the model had finished writing it. |
| 23 | |
| 24 | use serde_json::Value; |
| 25 | |
| 26 | /// Maximum raw argument length we'll attempt to repair (1 MiB). |
| 27 | const MAX_ARG_LEN: usize = 1024 * 1024; |
| 28 | |
| 29 | #[derive(Debug, thiserror::Error)] |
| 30 | pub enum ArgRepairError { |
| 31 | #[error("argument exceeded {0} chars; refusing to repair")] |
| 32 | TooLarge(usize), |
| 33 | #[error("argument could not be repaired into valid JSON")] |
| 34 | Unrepairable, |
| 35 | } |
| 36 | |
| 37 | /// Repair a raw JSON argument string into a valid `serde_json::Value`. |
| 38 | /// |
| 39 | /// Runs the deterministic ladder; on success returns the parsed value. |
| 40 | /// A repaired value plus whether the repair had to invent structure. |
| 41 | #[derive(Debug, Clone)] |
| 42 | pub struct Repaired { |
| 43 | pub value: Value, |
| 44 | /// True when the text only parsed after closers were appended (stage 4) |
| 45 | /// or discarded (stage 5) — i.e. the argument text was structurally |
| 46 | /// incomplete. Callers making the *final* dispatch decision must treat |
| 47 | /// this as malformed input rather than executing it. |
| 48 | pub structure_synthesized: bool, |
| 49 | } |
| 50 | |
| 51 | impl Repaired { |
| 52 | fn intact(value: Value) -> Self { |
| 53 | Self { |
| 54 | value, |
| 55 | structure_synthesized: false, |
| 56 | } |
| 57 | } |
| 58 | fn synthesized(value: Value) -> Self { |
| 59 | Self { |
| 60 | value, |
| 61 | structure_synthesized: true, |
| 62 | } |
| 63 | } |
| 64 | } |
| 65 | |
| 66 | pub fn repair(raw: &str) -> Result<Repaired, ArgRepairError> { |
| 67 | // Stage 1: strict parse. Valid JSON is returned intact at any size: the |
| 68 | // size bound exists to cap the cost of the repair stages below, and a |
| 69 | // strict serde parse is what callers already fall back to on oversize |
| 70 | // input. Checking size first made every valid argument over the bound |
| 71 | // (a large `write`, a generated fixture) fail as malformed. |
| 72 | if let Ok(v) = serde_json::from_str(raw) { |
| 73 | return Ok(Repaired::intact(v)); |
| 74 | } |
| 75 | if raw.len() > MAX_ARG_LEN { |
| 76 | return Err(ArgRepairError::TooLarge(raw.len())); |
| 77 | } |
| 78 | // Stage 2: strip control chars inside strings |
| 79 | let mut s = strip_control_chars_in_strings(raw); |
| 80 | if let Ok(v) = serde_json::from_str(&s) { |
| 81 | return Ok(Repaired::intact(v)); |
| 82 | } |
| 83 | // Stage 3: strip trailing commas |
| 84 | s = strip_trailing_commas(&s); |
| 85 | if let Ok(v) = serde_json::from_str(&s) { |
| 86 | return Ok(Repaired::intact(v)); |
| 87 | } |
| 88 | // Stage 4: balance braces |
| 89 | // Stages 4 and 5 change structure. Anything they rescue is reported as |
| 90 | // synthesized: `balance_braces` counts braces without tracking string |
| 91 | // literals, so a stream cut at the end of a complete string value yields |
| 92 | // JSON that parses cleanly and is still missing whatever the model had |
| 93 | // not written yet. |
| 94 | s = balance_braces(&s, 50); |
| 95 | if let Ok(v) = serde_json::from_str(&s) { |
| 96 | return Ok(Repaired::synthesized(v)); |
| 97 | } |
| 98 | // Stage 5: strip excess closers |
| 99 | s = strip_excess_closers(&s); |
| 100 | if let Ok(v) = serde_json::from_str(&s) { |
| 101 | return Ok(Repaired::synthesized(v)); |
| 102 | } |
| 103 | Err(ArgRepairError::Unrepairable) |
| 104 | } |
| 105 | |
| 106 | /// Strip ASCII control characters (0x00–0x1F except \t, \n, \r) that appear |
| 107 | /// inside JSON string values. We walk character-by-character tracking whether |
| 108 | /// we're inside a string (between unescaped double-quotes). |
| 109 | fn strip_control_chars_in_strings(s: &str) -> String { |
| 110 | let mut out = String::with_capacity(s.len()); |
| 111 | let mut in_string = false; |
| 112 | let mut escape = false; |
| 113 | for ch in s.chars() { |
| 114 | if escape { |
| 115 | out.push(ch); |
| 116 | escape = false; |
| 117 | continue; |
| 118 | } |
| 119 | if ch == '\\' { |
| 120 | escape = true; |
| 121 | out.push(ch); |
| 122 | continue; |
| 123 | } |
| 124 | if ch == '"' { |
| 125 | in_string = !in_string; |
| 126 | out.push(ch); |
| 127 | continue; |
| 128 | } |
| 129 | if in_string && (ch as u32) < 0x20 && ch != '\t' && ch != '\n' && ch != '\r' { |
| 130 | // Drop control characters inside strings |
| 131 | continue; |
| 132 | } |
| 133 | out.push(ch); |
| 134 | } |
| 135 | out |
| 136 | } |
| 137 | |
| 138 | /// Strip trailing commas before `}` or `]` (optionally across whitespace) |
| 139 | /// and at end of input. String-aware: a `,}` or `,]` inside a string value |
| 140 | /// is content, e.g. source code in a `write` call, and is left untouched. |
| 141 | /// |
| 142 | /// Single pass: a run of commas and whitespace outside a string is held |
| 143 | /// back until the next significant character decides it, so a long run of |
| 144 | /// commas (a model repetition loop, re-repaired on every streamed delta) |
| 145 | /// costs linear time rather than a rescan per comma. |
| 146 | fn strip_trailing_commas(s: &str) -> String { |
| 147 | let mut out = String::with_capacity(s.len()); |
| 148 | // Held-back commas and whitespace, and whether it holds any comma. |
| 149 | let mut held = String::new(); |
| 150 | let mut held_comma = false; |
| 151 | let mut in_string = false; |
| 152 | let mut escape = false; |
| 153 | for ch in s.chars() { |
| 154 | if in_string { |
| 155 | if escape { |
| 156 | escape = false; |
| 157 | } else if ch == '\\' { |
| 158 | escape = true; |
| 159 | } else if ch == '"' { |
| 160 | in_string = false; |
| 161 | } |
| 162 | out.push(ch); |
| 163 | continue; |
| 164 | } |
| 165 | if ch == ',' { |
| 166 | held.push(ch); |
| 167 | held_comma = true; |
| 168 | continue; |
| 169 | } |
| 170 | if held_comma && ch.is_whitespace() { |
| 171 | held.push(ch); |
| 172 | continue; |
| 173 | } |
| 174 | if held_comma { |
| 175 | flush_held_commas(&mut out, &held, matches!(ch, '}' | ']')); |
| 176 | held.clear(); |
| 177 | held_comma = false; |
| 178 | } |
| 179 | if ch == '"' { |
| 180 | in_string = true; |
| 181 | } |
| 182 | out.push(ch); |
| 183 | } |
| 184 | if held_comma { |
| 185 | flush_held_commas(&mut out, &held, true); |
| 186 | } |
| 187 | out |
| 188 | } |
| 189 | |
| 190 | /// Emit a held comma/whitespace run: dropping its commas when it trails |
| 191 | /// before a closer or the end of input, keeping it verbatim otherwise. |
| 192 | fn flush_held_commas(out: &mut String, held: &str, drop_commas: bool) { |
| 193 | if drop_commas { |
| 194 | out.extend(held.chars().filter(|c| *c != ',')); |
| 195 | } else { |
| 196 | out.push_str(held); |
| 197 | } |
| 198 | } |
| 199 | |
| 200 | /// Balance braces and brackets: count `{`/`}` and `[`/`]`, append closers if |
| 201 | /// positive delta (more opens than closes). Caps iterations so a |
| 202 | /// catastrophically broken input doesn't loop forever. |
| 203 | fn balance_braces(s: &str, max_iter: usize) -> String { |
| 204 | let mut out = s.to_string(); |
| 205 | for _ in 0..max_iter { |
| 206 | let brace_delta: i32 = out |
| 207 | .chars() |
| 208 | .map(|ch| match ch { |
| 209 | '{' => 1, |
| 210 | '}' => -1, |
| 211 | _ => 0, |
| 212 | }) |
| 213 | .sum(); |
| 214 | let bracket_delta: i32 = out |
| 215 | .chars() |
| 216 | .map(|ch| match ch { |
| 217 | '[' => 1, |
| 218 | ']' => -1, |
| 219 | _ => 0, |
| 220 | }) |
| 221 | .sum(); |
| 222 | if brace_delta <= 0 && bracket_delta <= 0 { |
| 223 | break; |
| 224 | } |
| 225 | // Append needed closers in reverse order (brackets before braces |
| 226 | // for correct nesting when both are unbalanced). |
| 227 | for _ in 0..bracket_delta.max(0) { |
| 228 | out.push(']'); |
| 229 | } |
| 230 | for _ in 0..brace_delta.max(0) { |
| 231 | out.push('}'); |
| 232 | } |
| 233 | } |
| 234 | out |
| 235 | } |
| 236 | |
| 237 | /// Strip excess closers when the delta is negative (more closes than opens). |
| 238 | fn strip_excess_closers(s: &str) -> String { |
| 239 | let mut brace_depth: i32 = 0; |
| 240 | let mut bracket_depth: i32 = 0; |
| 241 | let mut out = String::with_capacity(s.len()); |
| 242 | for ch in s.chars() { |
| 243 | match ch { |
| 244 | '}' => { |
| 245 | if brace_depth > 0 { |
| 246 | brace_depth -= 1; |
| 247 | out.push(ch); |
| 248 | } |
| 249 | // else drop excess closer |
| 250 | } |
| 251 | ']' => { |
| 252 | if bracket_depth > 0 { |
| 253 | bracket_depth -= 1; |
| 254 | out.push(ch); |
| 255 | } |
| 256 | } |
| 257 | '{' => { |
| 258 | brace_depth += 1; |
| 259 | out.push(ch); |
| 260 | } |
| 261 | '[' => { |
| 262 | bracket_depth += 1; |
| 263 | out.push(ch); |
| 264 | } |
| 265 | _ => out.push(ch), |
| 266 | } |
| 267 | } |
| 268 | out |
| 269 | } |
| 270 | |
| 271 | #[cfg(test)] |
| 272 | mod tests { |
| 273 | use super::*; |
| 274 | use serde_json::json; |
| 275 | |
| 276 | #[test] |
| 277 | fn strict_parse_passes_through() { |
| 278 | let r = repair(r#"{"path": "hello.txt"}"#).unwrap(); |
| 279 | assert_eq!(r.value, json!({"path": "hello.txt"})); |
| 280 | assert!(!r.structure_synthesized); |
| 281 | } |
| 282 | |
| 283 | #[test] |
| 284 | fn repairs_trailing_comma() { |
| 285 | let r = repair(r#"{"path": "hello.txt",}"#).unwrap(); |
| 286 | assert_eq!(r.value, json!({"path": "hello.txt"})); |
| 287 | // Structurally complete, just sloppy — must stay dispatchable. |
| 288 | assert!(!r.structure_synthesized); |
| 289 | } |
| 290 | |
| 291 | #[test] |
| 292 | fn repairs_trailing_comma_in_array() { |
| 293 | let r = repair(r#"["a", "b",]"#).unwrap(); |
| 294 | assert_eq!(r.value, json!(["a", "b"])); |
| 295 | assert!(!r.structure_synthesized); |
| 296 | } |
| 297 | |
| 298 | #[test] |
| 299 | fn repairs_missing_close_brace() { |
| 300 | let r = repair(r#"{"path": "hello.txt""#).unwrap(); |
| 301 | assert_eq!(r.value, json!({"path": "hello.txt"})); |
| 302 | assert!(r.structure_synthesized); |
| 303 | } |
| 304 | |
| 305 | #[test] |
| 306 | fn repairs_missing_close_bracket() { |
| 307 | let r = repair(r#"["a", "b""#).unwrap(); |
| 308 | assert_eq!(r.value, json!(["a", "b"])); |
| 309 | assert!(r.structure_synthesized); |
| 310 | } |
| 311 | |
| 312 | #[test] |
| 313 | fn strips_embedded_control_chars() { |
| 314 | // Raw \x0B (vertical tab) inside a string value |
| 315 | let raw = "{\"key\": \"val\x0Bue\"}"; |
| 316 | let v = repair(raw).unwrap(); |
| 317 | assert_eq!(v.value, json!({"key": "value"})); |
| 318 | assert!(!v.structure_synthesized); |
| 319 | } |
| 320 | |
| 321 | #[test] |
| 322 | fn rejects_empty_string() { |
| 323 | assert!(matches!(repair(""), Err(ArgRepairError::Unrepairable))); |
| 324 | } |
| 325 | |
| 326 | #[test] |
| 327 | fn rejects_gibberish() { |
| 328 | assert!(matches!( |
| 329 | repair("not json at all"), |
| 330 | Err(ArgRepairError::Unrepairable) |
| 331 | )); |
| 332 | } |
| 333 | |
| 334 | #[test] |
| 335 | fn balances_nested_braces() { |
| 336 | let r = repair(r#"{"outer": {"inner": "val""#).unwrap(); |
| 337 | assert_eq!(r.value, json!({"outer": {"inner": "val"}})); |
| 338 | // Closers were appended, so this is a truncated argument, not a |
| 339 | // complete one that merely needed tidying. |
| 340 | assert!(r.structure_synthesized); |
| 341 | } |
| 342 | |
| 343 | #[test] |
| 344 | fn strips_excess_closers() { |
| 345 | let r = repair(r#"{"key": "val"}}"#).unwrap(); |
| 346 | assert_eq!(r.value, json!({"key": "val"})); |
| 347 | assert!(r.structure_synthesized); |
| 348 | } |
| 349 | |
| 350 | #[test] |
| 351 | fn handles_double_encoded_json() { |
| 352 | // This is a valid JSON string containing a JSON object literal. |
| 353 | // repair parses it as a string; the engine's existing fallback |
| 354 | // (parse_tool_input) will unwrap the string and re-parse. |
| 355 | let r = repair(r#""{\"path\": \"hello.txt\"}""#).unwrap(); |
| 356 | assert_eq!( |
| 357 | r.value, |
| 358 | Value::String(r#"{"path": "hello.txt"}"#.to_string()) |
| 359 | ); |
| 360 | assert!(!r.structure_synthesized); |
| 361 | } |
| 362 | |
| 363 | #[test] |
| 364 | fn oversize_input_rejected() { |
| 365 | let big = "x".repeat(MAX_ARG_LEN + 1); |
| 366 | assert!(repair(&big).is_err()); |
| 367 | } |
| 368 | |
| 369 | #[test] |
| 370 | fn oversize_valid_json_parses_intact() { |
| 371 | // A valid argument over the repair bound whose string content holds |
| 372 | // an unbalanced brace must come back exactly as written, not be |
| 373 | // refused as too large or "repaired". |
| 374 | let content = format!("fn main() {{{}", "x".repeat(MAX_ARG_LEN)); |
| 375 | let raw = serde_json::json!({"path": "big.rs", "content": content}).to_string(); |
| 376 | assert!(raw.len() > MAX_ARG_LEN); |
| 377 | let r = repair(&raw).expect("valid oversize JSON parses"); |
| 378 | assert!(!r.structure_synthesized); |
| 379 | assert_eq!(r.value["content"].as_str(), Some(content.as_str())); |
| 380 | |
| 381 | // Oversize text that is not valid JSON is still refused before any |
| 382 | // repair stage runs. |
| 383 | let truncated = &raw[..raw.len() - 2]; |
| 384 | assert!(matches!( |
| 385 | repair(truncated), |
| 386 | Err(ArgRepairError::TooLarge(_)) |
| 387 | )); |
| 388 | } |
| 389 | |
| 390 | #[test] |
| 391 | fn a_write_cut_at_a_string_boundary_is_reported_as_synthesized() { |
| 392 | // The defect this flag exists for: `balance_braces` counts braces |
| 393 | // without tracking string literals, so a provider that cuts the |
| 394 | // stream at its output limit right after a complete string value |
| 395 | // yields text that parses cleanly once one `}` is appended. Nothing |
| 396 | // downstream could previously tell this from a finished argument, so |
| 397 | // the truncated `content` was written to the user's file. |
| 398 | let cut = r#"{"path": "notes.md", "content": "first line""#; |
| 399 | let r = repair(cut).unwrap(); |
| 400 | assert_eq!( |
| 401 | r.value, |
| 402 | json!({"path": "notes.md", "content": "first line"}), |
| 403 | "the ladder still parses it — that is exactly why the flag is needed" |
| 404 | ); |
| 405 | assert!( |
| 406 | r.structure_synthesized, |
| 407 | "a truncated write must be reported as synthesized so dispatch refuses it" |
| 408 | ); |
| 409 | } |
| 410 | |
| 411 | #[test] |
| 412 | fn a_complete_argument_needing_only_control_char_stripping_stays_intact() { |
| 413 | // Stage 2 normalizes text that was already structurally complete, so |
| 414 | // it must NOT be flagged — otherwise every DeepSeek chunk-boundary |
| 415 | // repair would start failing tool calls that are perfectly fine. |
| 416 | let r = repair("{\"a\": \"line\u{0008}break\"}").unwrap(); |
| 417 | assert_eq!(r.value, json!({"a": "linebreak"})); |
| 418 | assert!(!r.structure_synthesized); |
| 419 | } |
| 420 | |
| 421 | #[test] |
| 422 | fn repairs_brace_balance_with_trailing_comma() { |
| 423 | let r = repair(r#"{"a": 1,"#).unwrap(); |
| 424 | assert_eq!(r.value, json!({"a": 1})); |
| 425 | assert!(r.structure_synthesized); |
| 426 | } |
| 427 | |
| 428 | #[test] |
| 429 | fn trailing_comma_repair_leaves_string_content_alone() { |
| 430 | // Object-level trailing comma forces the repair path; the `,}` and |
| 431 | // `,]` inside `content` are the model's code and must survive. |
| 432 | let raw = r#"{"path":"a.js","content":"const o = {a:1,};\nlet v = [1,2,];\n",}"#; |
| 433 | let r = repair(raw).unwrap(); |
| 434 | assert_eq!( |
| 435 | r.value, |
| 436 | json!({"path": "a.js", "content": "const o = {a:1,};\nlet v = [1,2,];\n"}) |
| 437 | ); |
| 438 | assert!(!r.structure_synthesized); |
| 439 | } |
| 440 | |
| 441 | #[test] |
| 442 | fn a_long_comma_run_is_repaired_in_linear_time() { |
| 443 | // A repetition loop of commas outside any string must not cost a |
| 444 | // rescan per comma: repair runs on every streamed partial buffer. |
| 445 | let mut raw = String::from(r#"{"a":[1"#); |
| 446 | for _ in 0..200_000 { |
| 447 | raw.push_str(", "); |
| 448 | } |
| 449 | raw.push_str("]}"); |
| 450 | let (tx, rx) = std::sync::mpsc::channel(); |
| 451 | std::thread::spawn(move || { |
| 452 | let _ = tx.send(repair(&raw).map(|r| r.value)); |
| 453 | }); |
| 454 | let repaired = rx |
| 455 | .recv_timeout(std::time::Duration::from_secs(5)) |
| 456 | .expect("trailing-comma repair must finish in linear time"); |
| 457 | assert_eq!(repaired.unwrap(), json!({"a": [1]})); |
| 458 | } |
| 459 | |
| 460 | #[test] |
| 461 | fn interior_comma_runs_are_left_for_the_parser_to_reject() { |
| 462 | // Only commas that trail before a closer are removed; a doubled |
| 463 | // separator between values is not invented away. |
| 464 | assert!(repair(r#"{"a":[1,,2]}"#).is_err()); |
| 465 | let r = repair("{\"a\":[1 , ,\n]}").unwrap(); |
| 466 | assert_eq!(r.value, json!({"a": [1]})); |
| 467 | assert!(!r.structure_synthesized); |
| 468 | } |
| 469 | |
| 470 | #[test] |
| 471 | fn trailing_comma_before_whitespace_and_closer_is_stripped() { |
| 472 | let r = repair("{\"a\": [1, 2 ,\n ],\n}").unwrap(); |
| 473 | assert_eq!(r.value, json!({"a": [1, 2]})); |
| 474 | assert!(!r.structure_synthesized); |
| 475 | } |
| 476 | } |
| 477 |