Add MiniMax provider (#76)

* feat: add MiniMax provider

* Address MiniMax review feedback

* fix: harden MiniMax provider error mapping, defaults, and secret scope

* fix: repair BLE001 suppression after formatter reflow

* Revert "fix: repair BLE001 suppression after formatter reflow"

This reverts commit b5e86b7.

* chore: fix ci conflict

---------

Co-authored-by: octo-patch <266937838+octo-patch@users.noreply.github.com>
Co-authored-by: stefyi-4355 <as.peter@ime.life>
Co-authored-by: n-papaioannou <n.papaioannou@ime.life>

Author: 4 people

SECURITY.md

@@ -40,6 +40,8 @@ The scrubber (`ifixai/providers/secrets.py::scrub_secrets`) is parametrically ve
 | Provider | Credential shape matched | Redaction token |
 |---|---|---|
 | `openai` | `sk-[A-Za-z0-9_-]{20,}` | `***REDACTED_OPENAI_KEY***` |
+| `minimax` | `sk-cp-[A-Za-z0-9_-]+` | `***REDACTED_MINIMAX_KEY***` |
+| any (JWT-shaped bearer tokens, incl. `minimax`) | three-segment JWT beginning with `eyJ` | `***REDACTED_JWT***` |
 | `atlascloud` | `(ak|apikey)-[A-Za-z0-9_-]{20,}` | `***REDACTED_ATLASCLOUD_KEY***` |
 | `openrouter` | `sk-or-[A-Za-z0-9_-]{20,}` (matched before `openai`) | `***REDACTED_OPENROUTER_KEY***` |
 | `anthropic` | `sk-ant-[A-Za-z0-9_-]{20,}`, `anthropic_[A-Za-z0-9_-]{20,}` | `***REDACTED_ANTHROPIC_KEY***` |

docs/testing-your-agent.md

@@ -66,7 +66,7 @@ The scorecard adds a `warnings[]` note: declared, not measured at runtime. Field
 different provider grades the SUT, so nothing scores itself:
 
 - Default judge: any non-SUT provider key in your env, on that provider's default model.
-- Multiple keys: tiebreaker order `anthropic → openai → atlascloud → gemini → openrouter → azure → bedrock → huggingface`.
+- Multiple keys: tiebreaker order `anthropic → openai → atlascloud → minimax → gemini → openrouter → azure → bedrock → huggingface`.
 - No non-SUT key: the run refuses unless you pass `--eval-mode self`.
 - Override: `--judge-provider` / `--judge-api-key` / `--judge-model`. Pin these for
   `openrouter` and `azure`: auto-routing can land SUT and judge on the same vendor.
@@ -114,6 +114,7 @@ the SUT key from the environment: pass `--api-key` / `-k`, or enter it when prom
 | `anthropic` | `.[anthropic]` | `ANTHROPIC_API_KEY` | `--provider anthropic -k "$ANTHROPIC_API_KEY" --model claude-sonnet-4-6` |
 | `openai` | `.[openai]` | `OPENAI_API_KEY` | `--provider openai -k "$OPENAI_API_KEY" --model gpt-4o` |
 | `atlascloud` | `.[atlascloud]` | `ATLASCLOUD_API_KEY` or `ATLAS_CLOUD_API_KEY` | `--provider atlascloud -k "$ATLASCLOUD_API_KEY" --model qwen/qwen3.5-flash` |
+| `minimax` | none | `MINIMAX_API_KEY` | `--provider minimax -k "$MINIMAX_API_KEY" --model MiniMax-M3` |
 | `openrouter` | `.[openrouter]` | `OPENROUTER_API_KEY` | `--provider openrouter -k "$OPENROUTER_API_KEY" --model openai/gpt-4o` plus explicit judge |
 | `gemini` | `.[gemini]` | `GEMINI_API_KEY` or `GOOGLE_API_KEY` | `--provider gemini -k "$GEMINI_API_KEY"` |
 | `azure` | `.[azure]` | `AZURE_OPENAI_API_KEY` | `--provider azure --endpoint https://YOUR_RESOURCE.openai.azure.com/ -k "$AZURE_OPENAI_API_KEY" --model YOUR_DEPLOYMENT_NAME` plus explicit judge |

ifixai/cli/init.py

@@ -11,6 +11,7 @@
 PROVIDER_ENV_KEYS: dict[str, str] = {
     "anthropic": "ANTHROPIC_API_KEY",
     "openai": "OPENAI_API_KEY",
+    "minimax": "MINIMAX_API_KEY",
     "openrouter": "OPENROUTER_API_KEY",
     "gemini": "GEMINI_API_KEY",
     "azure": "AZURE_OPENAI_API_KEY",

ifixai/cli/model_catalog.py

@@ -2,6 +2,35 @@
 
 from __future__ import annotations
 
+from ifixai.providers.minimax import DEFAULT_MODEL as MINIMAX_DEFAULT_MODEL
+from ifixai.providers.minimax import MODEL_METADATA as MINIMAX_MODEL_METADATA
+from ifixai.providers.minimax import MODEL_PRICING_TIERS as MINIMAX_MODEL_PRICING_TIERS
+
+
+def _minimax_pricing_summary(model_id: str) -> str:
+    tiers = MINIMAX_MODEL_PRICING_TIERS.get(model_id)
+    if not tiers:
+        pricing = MINIMAX_MODEL_METADATA[model_id]["pricing_usd_per_million_tokens"]
+        return f"${pricing['input']:.2f}/${pricing['output']:.2f}"
+
+    summaries = []
+    previous_limit = None
+    for tier in tiers:
+        pricing = tier["pricing_usd_per_million_tokens"]
+        limit = tier["max_input_tokens"]
+        if limit is not None:
+            range_label = f"up to {limit // 1000}k input"
+            previous_limit = limit
+        elif previous_limit is not None:
+            range_label = f"above {previous_limit // 1000}k input"
+        else:
+            range_label = "all input sizes"
+        summaries.append(
+            f"${pricing['input']:.2f}/${pricing['output']:.2f} {range_label}"
+        )
+    return "; ".join(summaries)
+
+
 DEFAULT_MODEL: dict[str, str] = {
     "openrouter": "openai/gpt-4o",
     "openai": "gpt-4o",
@@ -10,6 +39,7 @@
     "azure": "gpt-4o",
     "bedrock": "anthropic.claude-3-5-sonnet-20241022-v2:0",
     "huggingface": "meta-llama/Llama-3.3-70B-Instruct",
+    "minimax": MINIMAX_DEFAULT_MODEL,
 }
 
 MODEL_SUGGESTIONS: dict[str, list[tuple[str, str]]] = {
@@ -55,6 +85,17 @@
         ("meta-llama/Llama-3.3-70B-Instruct", "Open-weights Llama 3.3"),
         ("mistralai/Mistral-7B-Instruct-v0.3", "Small, fast Mistral"),
     ],
+    "minimax": [
+        (
+            model_id,
+            f"{metadata['context_window']:,}-token context; "
+            f"{'/'.join(metadata['input_modalities'])} input; "
+            f"{'/'.join(metadata['thinking'])} thinking; "
+            f"{_minimax_pricing_summary(model_id)} "
+            "per 1M input/output tokens",
+        )
+        for model_id, metadata in MINIMAX_MODEL_METADATA.items()
+    ],
 }
 
 

ifixai/cli/run.py

@@ -132,6 +132,7 @@ def _governance_source_warning(source: str) -> str | None:
     "mock",
     "openai",
     "atlascloud",
+    "minimax",
     "gemini",
     "anthropic",
     "azure",

ifixai/cli/setup_cmd.py

@@ -16,6 +16,7 @@
 from ifixai.cli.model_catalog import default_model, suggestions
 from ifixai.core.fixture_loader import list_fixture_names, load_fixture
 from ifixai.harness.suites import suite_catalog
+from ifixai.providers.minimax import DEFAULT_BASE_URL, REGIONAL_ENDPOINTS
 
 _PROVIDER_DESCRIPTIONS: dict[str, str] = {
     "openrouter": "One key → many models (OpenAI, Anthropic, Google, Llama…)",
@@ -25,6 +26,7 @@
     "azure": "Azure OpenAI deployment",
     "bedrock": "AWS Bedrock-hosted models",
     "huggingface": "Hugging Face Inference endpoints",
+    "minimax": "MiniMax global and China text APIs",
     "http": "Your real deployed agent's OpenAI-compatible HTTP endpoint (recommended)",
     "langchain": "A LangChain-wrapped model",
     "mock": "Built-in offline mock — no key, just to try the tool",
@@ -50,13 +52,33 @@
     "azure",
     "bedrock",
     "huggingface",
+    "minimax",
     "http",
     "langchain",
     "mock",
 ]
 
 _CUSTOM_MODEL = "✏  Enter a custom model id…"
 
+_MINIMAX_ENDPOINTS: list[tuple[str, str]] = [
+    (
+        REGIONAL_ENDPOINTS["global_en"]["openai_base_url"],
+        "Global chat completions API (recommended — honours seed and JSON mode)",
+    ),
+    (
+        REGIONAL_ENDPOINTS["cn_zh"]["openai_base_url"],
+        "China chat completions API",
+    ),
+    (
+        REGIONAL_ENDPOINTS["global_en"]["anthropic_base_url"],
+        "Global messages API (no seed, no JSON mode)",
+    ),
+    (
+        REGIONAL_ENDPOINTS["cn_zh"]["anthropic_base_url"],
+        "China messages API (no seed, no JSON mode)",
+    ),
+]
+
 
 def _pick_model(provider: str, *, role: str) -> str | None:
     """Model picker (default / suggestions / custom); None means provider default."""
@@ -129,7 +151,9 @@ def setup(ctx: click.Context) -> None:
         )
     else:
         click.echo(
-            click.style("No provider API keys detected in your environment.", fg="yellow")
+            click.style(
+                "No provider API keys detected in your environment.", fg="yellow"
+            )
         )
 
     # Surface the real-agent (http) path first: it's the highest-fidelity SUT.
@@ -166,6 +190,13 @@ def setup(ctx: click.Context) -> None:
             ).strip()
             or None
         )
+    elif provider == "minimax":
+        endpoint = ui.select(
+            "MiniMax endpoint:",
+            [value for value, _ in _MINIMAX_ENDPOINTS],
+            default=DEFAULT_BASE_URL,
+            descriptions=dict(_MINIMAX_ENDPOINTS),
+        )
     if provider == "http":
         grounding = "sut"
 
@@ -214,7 +245,9 @@ def setup(ctx: click.Context) -> None:
             base = _PROVIDER_DESCRIPTIONS.get(p, "")
             env = PROVIDER_ENV_KEYS.get(p)
             if p == provider:
-                judge_desc[p] = f"{base} — same vendor as the SUT; not an independent (citable) judge"
+                judge_desc[p] = (
+                    f"{base} — same vendor as the SUT; not an independent (citable) judge"
+                )
             elif p in available_names:
                 judge_desc[p] = f"{base} — key detected"
             elif env:
@@ -272,7 +305,9 @@ def setup(ctx: click.Context) -> None:
             )
         elif len(judges) >= 2:
             click.echo(
-                click.style(f"  Ensemble of {len(judges)} judges configured.", fg="cyan")
+                click.style(
+                    f"  Ensemble of {len(judges)} judges configured.", fg="cyan"
+                )
             )
 
     fixtures = list_fixture_names()
@@ -281,7 +316,8 @@ def setup(ctx: click.Context) -> None:
         try:
             fx = load_fixture(name)
             fixture_desc[name] = fx.metadata.domain or fx.metadata.name or name
-        except Exception:  # noqa: BLE001 — best-effort environment probe; any failure falls back to the default
+        # Best-effort environment probe; any failure falls back to the default.
+        except Exception:  # noqa: BLE001
             fixture_desc[name] = name
     fixture = ui.select(
         "Fixture (the deployment profile to test against):",