tool format hardening

bantam.pl

@@ -94,10 +94,27 @@ sub shell_exec {
   return trim($res) . "\n\nexit: $code";
 }
 
+sub sanitize_msgs {
+  my ($msgs) = @_;
+  return unless $msgs && ref($msgs) eq 'ARRAY';
+  for my $m (@$msgs) {
+    next unless ref($m) eq 'HASH' && ($m->{role} // '') eq 'assistant' && $m->{tool_calls};
+    for my $tc (@{$m->{tool_calls}}) {
+      next unless ref($tc) eq 'HASH' && $tc->{function};
+      my $astr = $tc->{function}{arguments} // '{}';
+      my $a = eval { decode_json($astr) };
+      if ($@ || ref($a) ne 'HASH') {
+        $tc->{function}{arguments} = encode_json({ invalid_raw => $astr // '' });
+      }
+    }
+  }
+}
+
 my @fib = (1, 1, 2, 3, 5, 8, 13, 21, 34);
 
 sub llm {
   my ($cfg, $msgs, $tools) = @_;
+  sanitize_msgs($msgs);
   my $ep = $cfg->{endpoint} =~ s/\/+$//r;
   my $url = "$ep/chat/completions";
   my %h = ('Content-Type' => 'application/json', 'User-Agent' => 'Mozilla/5.0 (compatible; Bantam/1.0)');
@@ -119,8 +136,10 @@ sub llm {
         my $d = eval { decode_json($res->{content}) };
         return $d->{choices}[0]{message} if $d && $d->{choices} && @{$d->{choices}};
       }
-      my $err = $res->{reason} || $res->{content} || "HTTP status $res->{status}";
+      my $status = $res->{status} // 0;
+      my $err = $res->{reason} || $res->{content} || "HTTP status $status";
       print "\r\e[K" if $_col; STDOUT->flush();
+      die "[HTTP error: $err]\n" if $status >= 400 && $status < 500 && $status != 408 && $status != 429;
       if ($i <= $#fib) { print c("[network error: $err, retrying in ${dly}s...]", 31), "\n"; sleep($dly); next; }
       die "[network error: $err]\n";
     } else {
@@ -175,8 +194,10 @@ sub llm {
         $msg{tool_calls} = [map { $tcs{$_} } @order] if @order;
         return \%msg;
       }
-      my $err = $res->{reason} || $res->{content} || "HTTP status $res->{status}";
+      my $status = $res->{status} // 0;
+      my $err = $res->{reason} || $res->{content} || "HTTP status $status";
       print "\r\e[K" if $_col; STDOUT->flush();
+      die "[HTTP error: $err]\n" if $status >= 400 && $status < 500 && $status != 408 && $status != 429;
       if ($i <= $#fib) { print c("[network error: $err, retrying in ${dly}s...]", 31), "\n"; sleep($dly); next; }
       die "[network error: $err]\n";
     }
@@ -208,6 +229,7 @@ sub AL {
       my ($res, $sty) = ('', 2);
       my $a = eval { decode_json($astr) };
       if ($@ || ref($a) ne 'HASH') {
+        $tc->{function}{arguments} = encode_json({ invalid_raw => $astr });
         $res = "[tool error: invalid JSON args for $fn: " . ($@ || 'not a JSON object') . ". Raw: '$astr']"; $sty = 31;
       } elsif ($fn eq 'shell_exec') {
         $res = shell_exec($a->{command} // '', $stime);

bantam.py

@@ -71,9 +71,25 @@ def shell(cmd, t=120):
     if isinstance(err, bytes): err = err.decode("utf-8", "replace")
     return f"{(out + err).strip()}\n\n[shell timeout after {t}s]\nexit: -1"
 
+def sanitize_msgs(msgs):
+  for m in msgs:
+    if isinstance(m, dict) and m.get("role") == "assistant" and "tool_calls" in m:
+      tcs = m.get("tool_calls") or []
+      for tc in tcs:
+        if isinstance(tc, dict) and "function" in tc:
+          fn = tc.get("function") or {}
+          astr = fn.get("arguments", "{}")
+          try:
+            p = json.loads(astr) if astr else {}
+            if not isinstance(p, dict):
+              fn["arguments"] = json.dumps({"invalid_raw": astr} if astr else {})
+          except Exception:
+            fn["arguments"] = json.dumps({"invalid_raw": astr} if astr else {})
+
 fib = [1, 1, 2, 3, 5, 8, 13, 21, 34]
 
 def llm(cfg, msgs, tools):
+  sanitize_msgs(msgs)
   url = cfg["endpoint"].rstrip("/") + "/chat/completions"
   h = {"Content-Type": "application/json", "User-Agent": "Mozilla/5.0 (compatible; Bantam/1.0)"}
   k = cfg.get("api_key", "").strip()
@@ -126,6 +142,11 @@ def llm(cfg, msgs, tools):
         return m
     except Exception as e:
       if _COL: sys.stdout.write("\r\033[K"); sys.stdout.flush()
+      if isinstance(e, urllib.error.HTTPError):
+        code = e.code
+        body = e.read().decode("utf-8", "replace").strip() if hasattr(e, "read") else str(e)
+        if 400 <= code < 500 and code not in (408, 429):
+          raise RuntimeError(f"HTTP {code}: {body}")
       if i < len(fib): print(c(f"[network error: {e}, retrying in {dly}s...]", 31)); time.sleep(dly)
       else: raise
 
@@ -148,7 +169,9 @@ def AL(cfg, msgs, sp, depth=0):
         a = json.loads(astr) if astr else {}
         if not isinstance(a, dict): raise ValueError("args must be a JSON object")
         err = ""
-      except Exception as e: err, a = str(e), {}
+      except Exception as e:
+        err, a = str(e), {}
+        tc["function"]["arguments"] = json.dumps({"invalid_raw": astr} if astr else {})
       if err: res, sty = f"[tool error: invalid JSON args for {fn}: {err}. Raw: {astr!r}]", 31
       elif fn == "shell_exec": res, sty = shell(a.get("command", ""), stime), 2
       elif fn == "run_subagent":

main.go

@@ -157,7 +157,24 @@ type streamDelta struct {
   } `json:"choices"`
 }
 
+func sanitizeMessages(msgs []Message) {
+  for i := range msgs {
+    if msgs[i].Role == "assistant" && len(msgs[i].ToolCalls) > 0 {
+      for j := range msgs[i].ToolCalls {
+        tc := &msgs[i].ToolCalls[j]
+        astr := tc.Function.Arguments
+        var a map[string]any
+        if err := json.Unmarshal([]byte(astr), &a); err != nil || a == nil {
+          fixed, _ := json.Marshal(map[string]string{"invalid_raw": astr})
+          tc.Function.Arguments = string(fixed)
+        }
+      }
+    }
+  }
+}
+
 func llm(cfg *Cfg, msgs []Message, tools []map[string]any) (Message, error) {
+  sanitizeMessages(msgs)
   p := map[string]any{"model": cfg.Model, "temperature": cfg.Temperature, "messages": msgs, "stream": cfg.Stream}
   if tools != nil { p["tools"] = tools }
   body, _ := json.Marshal(p)
@@ -176,14 +193,21 @@ func llm(cfg *Cfg, msgs []Message, tools []map[string]any) (Message, error) {
     req.Header.Set("User-Agent", "Mozilla/5.0 (compatible; Bantam/1.0)")
     if cfg.APIKey != "" && cfg.APIKey != "-" { req.Header.Set("Authorization", "Bearer "+cfg.APIKey) }
     resp, err = client.Do(req)
+    var is4xxClientErr bool
     if err == nil && resp.StatusCode >= 400 {
       b, _ := io.ReadAll(io.LimitReader(resp.Body, 4096))
       resp.Body.Close()
       err = fmt.Errorf("HTTP %d: %s", resp.StatusCode, strings.TrimSpace(string(b)))
+      if resp.StatusCode < 500 && resp.StatusCode != 408 && resp.StatusCode != 429 {
+        is4xxClientErr = true
+      }
       resp = nil
     }
     if err == nil { break }
     if COL { fmt.Print("\r\033[K") }
+    if is4xxClientErr {
+      return Message{}, err
+    }
     if i < len(fib) {
       fmt.Println(c(fmt.Sprintf("[network error: %v, retrying in %ds...]", err, fib[i]), 31))
       time.Sleep(time.Duration(fib[i]) * time.Second)
@@ -301,6 +325,15 @@ func AL(cfg *Cfg, msgs []Message, sp string, depth int) ([]Message, error) {
   for i := 0; i < cfg.MaxALIterations && !done; i++ {
     m, err := llm(cfg, msgs, TOOLS)
     if err != nil { return msgs, err }
+    for j := range m.ToolCalls {
+      tc := &m.ToolCalls[j]
+      astr := tc.Function.Arguments
+      var a map[string]any
+      if err := json.Unmarshal([]byte(astr), &a); err != nil || a == nil {
+        fixed, _ := json.Marshal(map[string]string{"invalid_raw": astr})
+        tc.Function.Arguments = string(fixed)
+      }
+    }
     msgs = append(msgs, m)
     if !cfg.Stream {
       if m.ReasoningContent != "" {

mb

@@ -4,13 +4,9 @@
 
 use strict; use warnings; use HTTP::Tiny; use JSON::PP; use POSIX qw(strftime); use File::Path qw(make_path);
 
-$SIG{__WARN__} = sub {
+$SIG{__WARN__} = sub { warn $_[0] unless $_[0] =~ /^Use of uninitialized value \$err in numeric eq \(==\) at .*IO\/Socket\/IP\.pm line \d+/ };
-  my $m = shift;
-  return if $m =~ /^Use of uninitialized value \$err in numeric eq \(==\) at .*IO\/Socket\/IP\.pm line \d+/;
-  warn $m;
-};
 
-binmode(STDOUT, ':encoding(UTF-8)');binmode(STDERR, ':encoding(UTF-8)');binmode(STDIN, ':encoding(UTF-8)');
+binmode(STDOUT, ':encoding(UTF-8)'); binmode(STDERR, ':encoding(UTF-8)'); binmode(STDIN, ':encoding(UTF-8)');
 
 $| = 1; # unbuffered output: live spinner, responsive prompt
 
@@ -29,7 +25,10 @@ sub sp { my $p = '';
 
 sub T { my ($n, $d, $p) = @_; {type=>'function', function=>{name=>$n, description=>$d, parameters=>{type=>'object', properties=>$p, required=>[keys %$p]}}} }
 
+sub sanitize_msgs { for my $m (@{$_[0]}) { if (ref $m eq 'HASH' && $m->{tool_calls}) { for my $tc (@{$m->{tool_calls}}) { my $a = eval { decode_json($tc->{function}{arguments} // '{}') }; $tc->{function}{arguments} = encode_json({invalid_raw => $tc->{function}{arguments} // ''}) if ref $a ne 'HASH'; } } } }
+
 sub llm { my ($c, $msgs) = @_; # one non-streaming chat completion
+  sanitize_msgs($msgs);
   my $ep = $c->{endpoint}; $ep =~ s{/+$}{};
   my $h = {'Content-Type'=>'application/json', 'User-Agent'=>'Mozilla/5.0 (compatible; MicroBantam/1.0)'};
   $h->{Authorization} = "Bearer $c->{api_key}" if $c->{api_key} ne '-';
@@ -66,11 +65,9 @@ sub AL { my ($c, $msgs, $sp, $depth) = @_; # the agentic loop: LLM <-> tools unt
       my $fn = $tc->{function}{name};
       my $a = eval { decode_json($tc->{function}{arguments} // '{}') };
       my $res;
-      if (ref $a ne 'HASH') { $res = "bad JSON args for $fn: $tc->{function}{arguments}"; }
+      if (ref $a ne 'HASH') { $tc->{function}{arguments} = encode_json({invalid_raw => $tc->{function}{arguments} // ''}); $res = "bad JSON args for $fn: " . ($tc->{function}{arguments} // ''); }
       elsif ($fn eq 'shell_exec') { $res = shell_exec($a->{command} // '', $c->{shell_timeout}); }
-      elsif ($fn eq 'run_subagent') { $res = $depth >= 5 ? '[subagent depth limit (5) reached, child not spawned]'
+      elsif ($fn eq 'run_subagent') { $res = $depth >= 5 ? '[subagent depth limit (5) reached, child not spawned]' : last_assistant(AL($c, [{role=>'system', content=>"$sp\n\nImportant: this is a child agent"}, {role=>'user', content=>$a->{prompt} // ''}], $sp, $depth + 1)); }
-        : last_assistant(AL($c, [{role=>'system', content=>"$sp\n\nImportant: this is a child agent"},
-                                 {role=>'user', content=>$a->{prompt} // ''}], $sp, $depth + 1)); }
       else { $res = "unknown tool: $fn"; }
       print "[tool] $fn: $res\n";
       push @$msgs, {role=>'tool', tool_call_id=>$tc->{id}, content=>$res};
@@ -86,8 +83,7 @@ sub save { sdir(); my $id = strftime('%Y%m%d-%H%M%S', localtime); my $i = 0;
   $id .= '-' . ++$i while -f "$SDIR/$id.json";
   open my $f, '>', "$SDIR/$id.json" or die "cannot save: $!";
   print $f JSON::PP->new->utf8->pretty->encode({id=>$id, messages=>$_[0]}); close $f; $id; }
-sub load { my ($want) = @_; my ($hit) = grep { $_->{id} eq $want } sessions();
+sub load { my ($want) = @_; my ($hit) = grep { $_->{id} eq $want } sessions(); die "no session: $want\n" unless $hit; $hit->{messages}; }
-  die "no session: $want\n" unless $hit; $hit->{messages}; }
 sub autosave { sdir(); open my $f, '>', "$SDIR/autosave.json" or return; print $f JSON::PP->new->utf8->pretty->encode({id=>'autosave', messages=>$_[0]}); }
 sub list_sessions { map { [$_->{id}, scalar @{$_->{messages} // []}] } sessions() }
 
@@ -95,14 +91,11 @@ sub main {
   my ($c, $sp) = (cfg(), sp());
   my $msgs = [{role=>'system', content=>$sp}];
   if (@ARGV) { open my $f, '<:encoding(UTF-8)', $ARGV[0] or die "cannot open $ARGV[0]: $!"; # file mode
-    local $/; push @$msgs, {role=>'user', content=><$f>};
+    local $/; push @$msgs, {role=>'user', content=><$f>}; AL($c, $msgs, $sp); autosave($msgs); return; }
-    AL($c, $msgs, $sp); autosave($msgs); return; }
   print "MicroBantam ready ($c->{model}). Commands: /quit /clear /save /list /load <id> /help\n";
   while (1) {
-    print "> "; my $u = <STDIN>;
+    print "> "; my $u = <STDIN>; last unless defined $u;
-    last unless defined $u;
+    $u =~ s/^\s+|\s+$//g; next unless length $u;
-    $u =~ s/^\s+|\s+$//g;
-    next unless length $u;
     if    ($u eq '/quit')  { last; }
     elsif ($u eq '/clear') { $msgs = [{role=>'system', content=>$sp}]; autosave($msgs); }
     elsif ($u eq '/save')  { print "session saved: ", save($msgs), "\n"; }