axiom 0.27.0 → 0.29.0

package/dist/bin.cjs

@@ -249,290 +249,7 @@ var import_api5 = require("@opentelemetry/api");
 
 // src/otel/semconv/attributes.ts
 var import_semantic_conventions = require("@opentelemetry/semantic-conventions");
-
-// src/otel/semconv/eval_proposal.ts
-var ATTR_EVAL_ID = "eval.id";
-var ATTR_EVAL_NAME = "eval.name";
-var ATTR_EVAL_VERSION = "eval.version";
-var ATTR_EVAL_TYPE = "eval.type";
-var ATTR_EVAL_TAGS = "eval.tags";
-var ATTR_EVAL_BASELINE_ID = "eval.baseline.id";
-var ATTR_EVAL_BASELINE_NAME = "eval.baseline.name";
-var ATTR_EVAL_BASELINE_VERSION = "eval.baseline.version";
-var ATTR_EVAL_METADATA = "eval.metadata";
-var ATTR_EVAL_CAPABILITY_NAME = "eval.capability.name";
-var ATTR_EVAL_STEP_NAME = "eval.step.name";
-var ATTR_EVAL_COLLECTION_ID = "eval.collection.id";
-var ATTR_EVAL_COLLECTION_SIZE = "eval.collection.size";
-var ATTR_EVAL_COLLECTION_NAME = "eval.collection.name";
-var ATTR_EVAL_CONFIG_FLAGS = "eval.config.flags";
-var ATTR_EVAL_CASE_INDEX = "eval.case.index";
-var ATTR_EVAL_CASE_INPUT = "eval.case.input";
-var ATTR_EVAL_CASE_OUTPUT = "eval.case.output";
-var ATTR_EVAL_CASE_EXPECTED = "eval.case.expected";
-var ATTR_EVAL_CASE_SCORES = "eval.case.scores";
-var ATTR_EVAL_CASE_METADATA = "eval.case.metadata";
-var ATTR_EVAL_TASK_OUTPUT = "eval.task.output";
-var ATTR_EVAL_TASK_NAME = "eval.task.name";
-var ATTR_EVAL_TASK_TYPE = "eval.task.type";
-var ATTR_EVAL_RUN_ID = "eval.run.id";
-var ATTR_EVAL_SCORE_NAME = "eval.score.name";
-var ATTR_EVAL_SCORE_VALUE = "eval.score.value";
-var ATTR_EVAL_SCORE_THRESHOLD = "eval.score.threshold";
-var ATTR_EVAL_SCORE_PASSED = "eval.score.passed";
-var ATTR_EVAL_SCORE_METADATA = "eval.score.metadata";
-var ATTR_EVAL_USER_NAME = "eval.user.name";
-var ATTR_EVAL_USER_EMAIL = "eval.user.email";
-
-// src/otel/semconv/attributes.ts
 var import_incubating = require("@opentelemetry/semantic-conventions/incubating");
-var ATTR_AXIOM_GEN_AI_SCHEMA_URL = "axiom.gen_ai.schema_url";
-var ATTR_AXIOM_GEN_AI_SDK_NAME = "axiom.gen_ai.sdk.name";
-var ATTR_AXIOM_GEN_AI_SDK_VERSION = "axiom.gen_ai.sdk.version";
-var ATTR_GEN_AI_CAPABILITY_NAME = "gen_ai.capability.name";
-var ATTR_GEN_AI_STEP_NAME = "gen_ai.step.name";
-var ATTR_GEN_AI_TOOL_ARGUMENTS = "gen_ai.tool.arguments";
-var ATTR_GEN_AI_TOOL_MESSAGE = "gen_ai.tool.message";
-var GEN_AI_PROVIDER_NAME_VALUE_ASSEMBLYAI = "assemblyai";
-var GEN_AI_PROVIDER_NAME_VALUE_CEREBRAS = "cerebras";
-var GEN_AI_PROVIDER_NAME_VALUE_DEEPGRAM = "deepgram";
-var GEN_AI_PROVIDER_NAME_VALUE_DEEPINFRA = "deepinfra";
-var GEN_AI_PROVIDER_NAME_VALUE_ELEVENLABS = "elevenlabs";
-var GEN_AI_PROVIDER_NAME_VALUE_FAL = "fal";
-var GEN_AI_PROVIDER_NAME_VALUE_FIREWORKS = "fireworks";
-var GEN_AI_PROVIDER_NAME_VALUE_GLADIA = "gladia";
-var GEN_AI_PROVIDER_NAME_VALUE_HUME = "hume";
-var GEN_AI_PROVIDER_NAME_VALUE_LMNT = "lmnt";
-var GEN_AI_PROVIDER_NAME_VALUE_LUMA = "luma";
-var GEN_AI_PROVIDER_NAME_VALUE_REPLICATE = "replicate";
-var GEN_AI_PROVIDER_NAME_VALUE_REVAI = "revai";
-var GEN_AI_PROVIDER_NAME_VALUE_TOGETHERAI = "togetherai";
-var GEN_AI_PROVIDER_NAME_VALUE_VERCEL = "vercel";
-var Attr = {
-  __EXPERIMENTAL_Flag: (flagName) => `flag.${flagName}`,
-  __EXPERIMENTAL_Fact: (factName) => `fact.${factName}`,
-  Axiom: {
-    GenAI: {
-      SchemaURL: ATTR_AXIOM_GEN_AI_SCHEMA_URL,
-      SDK: {
-        Name: ATTR_AXIOM_GEN_AI_SDK_NAME,
-        Version: ATTR_AXIOM_GEN_AI_SDK_VERSION
-      }
-    }
-  },
-  GenAI: {
-    PromptMetadata: {
-      ID: "axiom.gen_ai.prompt.id",
-      Name: "axiom.gen_ai.prompt.name",
-      Slug: "axiom.gen_ai.prompt.slug",
-      Version: "axiom.gen_ai.prompt.version"
-    },
-    /**
-     * These two are used to identify the span
-     */
-    Capability: {
-      Name: ATTR_GEN_AI_CAPABILITY_NAME
-    },
-    Step: {
-      Name: ATTR_GEN_AI_STEP_NAME
-    },
-    Provider: {
-      Name: import_incubating.ATTR_GEN_AI_PROVIDER_NAME,
-      Name_Values: {
-        Anthropic: import_incubating.GEN_AI_PROVIDER_NAME_VALUE_ANTHROPIC,
-        AssemblyAI: GEN_AI_PROVIDER_NAME_VALUE_ASSEMBLYAI,
-        AWSBedrock: import_incubating.GEN_AI_PROVIDER_NAME_VALUE_AWS_BEDROCK,
-        AzureAIInference: import_incubating.GEN_AI_PROVIDER_NAME_VALUE_AZURE_AI_INFERENCE,
-        AzureAIOpenAI: import_incubating.GEN_AI_PROVIDER_NAME_VALUE_AZURE_AI_OPENAI,
-        Cerebras: GEN_AI_PROVIDER_NAME_VALUE_CEREBRAS,
-        Cohere: import_incubating.GEN_AI_PROVIDER_NAME_VALUE_COHERE,
-        Deepgram: GEN_AI_PROVIDER_NAME_VALUE_DEEPGRAM,
-        DeepInfra: GEN_AI_PROVIDER_NAME_VALUE_DEEPINFRA,
-        Deepseek: import_incubating.GEN_AI_PROVIDER_NAME_VALUE_DEEPSEEK,
-        ElevenLabs: GEN_AI_PROVIDER_NAME_VALUE_ELEVENLABS,
-        Fal: GEN_AI_PROVIDER_NAME_VALUE_FAL,
-        Fireworks: GEN_AI_PROVIDER_NAME_VALUE_FIREWORKS,
-        GCPGemini: import_incubating.GEN_AI_PROVIDER_NAME_VALUE_GCP_GEMINI,
-        GCPGenAI: import_incubating.GEN_AI_PROVIDER_NAME_VALUE_GCP_GEN_AI,
-        GCPVertexAI: import_incubating.GEN_AI_PROVIDER_NAME_VALUE_GCP_VERTEX_AI,
-        Gladia: GEN_AI_PROVIDER_NAME_VALUE_GLADIA,
-        Groq: import_incubating.GEN_AI_PROVIDER_NAME_VALUE_GROQ,
-        Hume: GEN_AI_PROVIDER_NAME_VALUE_HUME,
-        IBMWatsonxAI: import_incubating.GEN_AI_PROVIDER_NAME_VALUE_IBM_WATSONX_AI,
-        Lmnt: GEN_AI_PROVIDER_NAME_VALUE_LMNT,
-        Luma: GEN_AI_PROVIDER_NAME_VALUE_LUMA,
-        MistralAI: import_incubating.GEN_AI_PROVIDER_NAME_VALUE_MISTRAL_AI,
-        OpenAI: import_incubating.GEN_AI_PROVIDER_NAME_VALUE_OPENAI,
-        Perplexity: import_incubating.GEN_AI_PROVIDER_NAME_VALUE_PERPLEXITY,
-        Replicate: GEN_AI_PROVIDER_NAME_VALUE_REPLICATE,
-        RevAI: GEN_AI_PROVIDER_NAME_VALUE_REVAI,
-        TogetherAI: GEN_AI_PROVIDER_NAME_VALUE_TOGETHERAI,
-        Vercel: GEN_AI_PROVIDER_NAME_VALUE_VERCEL,
-        XAI: import_incubating.GEN_AI_PROVIDER_NAME_VALUE_X_AI
-      }
-    },
-    /**
-     * Regular attributes
-     */
-    Agent: {
-      Description: import_incubating.ATTR_GEN_AI_AGENT_DESCRIPTION,
-      // not yet used by axiom-ai
-      ID: import_incubating.ATTR_GEN_AI_AGENT_ID,
-      // not yet used by axiom-ai
-      Name: import_incubating.ATTR_GEN_AI_AGENT_NAME
-      // not yet used by axiom-ai
-    },
-    Conversation: {
-      ID: import_incubating.ATTR_GEN_AI_CONVERSATION_ID
-      // not yet used by axiom-ai, anyway probably needs to be provided by user
-    },
-    Input: {
-      Messages: import_incubating.ATTR_GEN_AI_INPUT_MESSAGES
-    },
-    Operation: {
-      Name: import_incubating.ATTR_GEN_AI_OPERATION_NAME,
-      Name_Values: {
-        /**
-         * Note that "text_completion" is deprecated in favor of "chat" for both OpenAI and Anthropic
-         */
-        Chat: import_incubating.GEN_AI_OPERATION_NAME_VALUE_CHAT,
-        CreateAgent: import_incubating.GEN_AI_OPERATION_NAME_VALUE_CREATE_AGENT,
-        Embeddings: import_incubating.GEN_AI_OPERATION_NAME_VALUE_EMBEDDINGS,
-        ExecuteTool: import_incubating.GEN_AI_OPERATION_NAME_VALUE_EXECUTE_TOOL,
-        GenerateContent: import_incubating.GEN_AI_OPERATION_NAME_VALUE_GENERATE_CONTENT,
-        InvokeAgent: import_incubating.GEN_AI_OPERATION_NAME_VALUE_INVOKE_AGENT
-      }
-    },
-    Output: {
-      Messages: import_incubating.ATTR_GEN_AI_OUTPUT_MESSAGES,
-      Type: import_incubating.ATTR_GEN_AI_OUTPUT_TYPE,
-      Type_Values: {
-        Text: import_incubating.GEN_AI_OUTPUT_TYPE_VALUE_TEXT,
-        Json: import_incubating.GEN_AI_OUTPUT_TYPE_VALUE_JSON,
-        Image: import_incubating.GEN_AI_OUTPUT_TYPE_VALUE_IMAGE,
-        Speech: import_incubating.GEN_AI_OUTPUT_TYPE_VALUE_SPEECH
-      }
-    },
-    /**
-     * The provider that is hosting the model, eg AWS Bedrock
-     * There doesn't seem to be a semconv for this
-     */
-    Request: {
-      ChoiceCount: import_incubating.ATTR_GEN_AI_REQUEST_CHOICE_COUNT,
-      // not yet used by axiom-ai
-      EncodingFormats: import_incubating.ATTR_GEN_AI_REQUEST_ENCODING_FORMATS,
-      // not yet used by axiom-ai
-      FrequencyPenalty: import_incubating.ATTR_GEN_AI_REQUEST_FREQUENCY_PENALTY,
-      MaxTokens: import_incubating.ATTR_GEN_AI_REQUEST_MAX_TOKENS,
-      /**
-       * The model you asked for
-       */
-      Model: import_incubating.ATTR_GEN_AI_REQUEST_MODEL,
-      PresencePenalty: import_incubating.ATTR_GEN_AI_REQUEST_PRESENCE_PENALTY,
-      Seed: import_incubating.ATTR_GEN_AI_REQUEST_SEED,
-      StopSequences: import_incubating.ATTR_GEN_AI_REQUEST_STOP_SEQUENCES,
-      Temperature: import_incubating.ATTR_GEN_AI_REQUEST_TEMPERATURE,
-      TopK: import_incubating.ATTR_GEN_AI_REQUEST_TOP_K,
-      TopP: import_incubating.ATTR_GEN_AI_REQUEST_TOP_P
-    },
-    Response: {
-      FinishReasons: import_incubating.ATTR_GEN_AI_RESPONSE_FINISH_REASONS,
-      ID: import_incubating.ATTR_GEN_AI_RESPONSE_ID,
-      /**
-       * The model that was actually used (might be different bc routing) - only ever get this from the response, otherwise omit
-       */
-      Model: import_incubating.ATTR_GEN_AI_RESPONSE_MODEL
-      // somehow not landing on the span for google models? check up on this...
-    },
-    Tool: {
-      CallID: import_incubating.ATTR_GEN_AI_TOOL_CALL_ID,
-      Description: import_incubating.ATTR_GEN_AI_TOOL_DESCRIPTION,
-      Name: import_incubating.ATTR_GEN_AI_TOOL_NAME,
-      Type: import_incubating.ATTR_GEN_AI_TOOL_TYPE,
-      /**
-       * Note, OTel Semantic Convention suggest only putting tool inputs/outputs on the parent chat span
-       * But we at least want to give users THE OPTION to put them on the tool spans themselves as well
-       * Because it enables a lot of things with querying
-       * @see https://github.com/open-telemetry/semantic-conventions/releases/tag/v1.37.0
-       */
-      Arguments: ATTR_GEN_AI_TOOL_ARGUMENTS,
-      /**
-       * Note, OTel Semantic Convention suggest only putting tool inputs/outputs on the parent chat span
-       * But we at least want to give users THE OPTION to put them on the tool spans themselves as well
-       * Because it enables a lot of things with querying
-       * @see https://github.com/open-telemetry/semantic-conventions/releases/tag/v1.37.0
-       */
-      Message: ATTR_GEN_AI_TOOL_MESSAGE
-    },
-    Usage: {
-      InputTokens: import_incubating.ATTR_GEN_AI_USAGE_INPUT_TOKENS,
-      OutputTokens: import_incubating.ATTR_GEN_AI_USAGE_OUTPUT_TOKENS
-    }
-  },
-  Eval: {
-    ID: ATTR_EVAL_ID,
-    Name: ATTR_EVAL_NAME,
-    Version: ATTR_EVAL_VERSION,
-    Type: ATTR_EVAL_TYPE,
-    Baseline: {
-      ID: ATTR_EVAL_BASELINE_ID,
-      Name: ATTR_EVAL_BASELINE_NAME,
-      Version: ATTR_EVAL_BASELINE_VERSION
-    },
-    Capability: {
-      Name: ATTR_EVAL_CAPABILITY_NAME
-    },
-    Step: {
-      Name: ATTR_EVAL_STEP_NAME
-    },
-    Tags: ATTR_EVAL_TAGS,
-    Metadata: ATTR_EVAL_METADATA,
-    Collection: {
-      ID: ATTR_EVAL_COLLECTION_ID,
-      Name: ATTR_EVAL_COLLECTION_NAME,
-      Size: ATTR_EVAL_COLLECTION_SIZE
-    },
-    Config: {
-      Flags: ATTR_EVAL_CONFIG_FLAGS
-    },
-    Run: {
-      ID: ATTR_EVAL_RUN_ID
-    },
-    Case: {
-      Index: ATTR_EVAL_CASE_INDEX,
-      Input: ATTR_EVAL_CASE_INPUT,
-      Output: ATTR_EVAL_CASE_OUTPUT,
-      Expected: ATTR_EVAL_CASE_EXPECTED,
-      Scores: ATTR_EVAL_CASE_SCORES,
-      Metadata: ATTR_EVAL_CASE_METADATA
-    },
-    Task: {
-      Output: ATTR_EVAL_TASK_OUTPUT,
-      Name: ATTR_EVAL_TASK_NAME,
-      Type: ATTR_EVAL_TASK_TYPE
-    },
-    Score: {
-      Name: ATTR_EVAL_SCORE_NAME,
-      Value: ATTR_EVAL_SCORE_VALUE,
-      Threshold: ATTR_EVAL_SCORE_THRESHOLD,
-      Passed: ATTR_EVAL_SCORE_PASSED,
-      Metadata: ATTR_EVAL_SCORE_METADATA
-    },
-    User: {
-      Name: ATTR_EVAL_USER_NAME,
-      Email: ATTR_EVAL_USER_EMAIL
-    }
-  },
-  Error: {
-    Type: import_semantic_conventions.ATTR_ERROR_TYPE,
-    Message: import_incubating.ATTR_ERROR_MESSAGE
-  },
-  HTTP: {
-    Response: {
-      StatusCode: import_semantic_conventions.ATTR_HTTP_RESPONSE_STATUS_CODE
-    }
-  }
-};
 
 // src/otel/startActiveSpan.ts
 var import_api2 = require("@opentelemetry/api");
@@ -543,7 +260,7 @@ var import_api4 = require("@opentelemetry/api");
 // package.json
 var package_default = {
   name: "axiom",
-  version: "0.27.0",
+  version: "0.29.0",
   type: "module",
   author: "Axiom, Inc.",
   contributors: [
@@ -747,191 +464,6 @@ function withEvalContext(options = {}, fn) {
   );
 }
 
-// src/config/resolver.ts
-var buildConsoleUrl = (urlString) => {
-  const url = new URL(urlString);
-  return `${url.protocol}//app.${url.host.split("api.").at(-1)}`;
-};
-function resolveAxiomConnection(config) {
-  let consoleEndpointUrl = buildConsoleUrl(config.eval.url);
-  if ("__overrideEndpointUrl" in config.eval) {
-    consoleEndpointUrl = config.eval.__overrideEndpointUrl;
-  }
-  return {
-    url: config.eval.url,
-    consoleEndpointUrl,
-    token: config.eval.token,
-    dataset: config.eval.dataset,
-    orgId: config.eval.orgId
-  };
-}
-
-// src/cli/errors.ts
-var AxiomCLIError = class extends Error {
-  constructor(message) {
-    super(message);
-    this.name = "AxiomCLIError";
-  }
-};
-function errorToString(error) {
-  if (typeof error === "string") {
-    return error;
-  }
-  if (error instanceof Error) {
-    return error.message;
-  }
-  return JSON.stringify(error);
-}
-
-// src/evals/eval.service.ts
-var findEvaluationCases = async (evalId, config) => {
-  const { dataset, url, token, orgId } = resolveAxiomConnection(config);
-  const apl = `['${dataset}'] | where trace_id == "${evalId}" | order by _time`;
-  const headers = new Headers({
-    Authorization: `Bearer ${token}`,
-    "Content-Type": "application/json",
-    ...orgId ? { "X-AXIOM-ORG-ID": orgId } : {}
-  });
-  const resp = await fetch(`${url}/v1/datasets/_apl?format=legacy`, {
-    headers,
-    method: "POST",
-    body: JSON.stringify({ apl })
-  });
-  const payload = await resp.json();
-  if (!resp.ok) {
-    throw new Error(`Failed to query evaluation cases: ${payload.message || resp.statusText}`);
-  }
-  return payload.matches.length ? buildSpanTree(payload.matches) : null;
-};
-var mapSpanToEval = (span) => {
-  const flagConfigRaw = span.data.attributes[Attr.Eval.Config.Flags] ?? span.data.attributes.custom[Attr.Eval.Config.Flags];
-  return {
-    id: span.data.attributes.custom[Attr.Eval.ID],
-    name: span.data.attributes.custom[Attr.Eval.Name],
-    type: span.data.attributes.custom[Attr.Eval.Type],
-    version: span.data.attributes.custom[Attr.Eval.Version],
-    collection: {
-      name: span.data.attributes.custom[Attr.Eval.Collection.Name],
-      size: span.data.attributes.custom[Attr.Eval.Collection.Size]
-    },
-    baseline: {
-      id: span.data.attributes.custom[Attr.Eval.Baseline.ID],
-      name: span.data.attributes.custom[Attr.Eval.Baseline.Name]
-    },
-    prompt: {
-      // TODO: do we still want this?
-      model: span.data.attributes.custom["eval.prompt.model"],
-      params: span.data.attributes.custom["eval.prompt.params"]
-    },
-    duration: span.data.duration,
-    status: span.data.status.code,
-    traceId: span.data.trace_id,
-    runAt: span._time,
-    tags: span.data.attributes.custom[Attr.Eval.Tags].length ? JSON.parse(span.data.attributes.custom[Attr.Eval.Tags]) : [],
-    user: {
-      name: span.data.attributes.custom[Attr.Eval.User.Name],
-      email: span.data.attributes.custom[Attr.Eval.User.Email]
-    },
-    cases: [],
-    flagConfig: flagConfigRaw ? JSON.parse(flagConfigRaw) : void 0
-  };
-};
-var mapSpanToCase = (item) => {
-  const data = item.data;
-  const d = data.duration;
-  let duration = "-";
-  if (d.endsWith("s")) {
-    duration = `${Number(d.replace("s", "")).toFixed(2)}s`;
-  } else {
-    duration = d;
-  }
-  return {
-    index: data.attributes.custom[Attr.Eval.Case.Index],
-    input: data.attributes.custom[Attr.Eval.Case.Input],
-    output: data.attributes.custom[Attr.Eval.Case.Output],
-    expected: data.attributes.custom[Attr.Eval.Case.Expected],
-    duration,
-    status: data.status.code,
-    scores: data.attributes.custom[Attr.Eval.Case.Scores] ? JSON.parse(data.attributes.custom[Attr.Eval.Case.Scores]) : {},
-    runAt: item._time,
-    spanId: data.span_id,
-    traceId: data.trace_id
-  };
-};
-var buildSpanTree = (spans) => {
-  if (!spans.length) {
-    return null;
-  }
-  const evalSpan = spans.find((span) => span.data.attributes.gen_ai.operation.name === "eval");
-  if (!evalSpan) {
-    return null;
-  }
-  const rootSpan = mapSpanToEval(evalSpan);
-  const caseSpans = spans.filter((span) => span.data.name.startsWith("case"));
-  for (const caseSpan of caseSpans) {
-    const caseData = mapSpanToCase(caseSpan);
-    const taskSpans = spans.filter(
-      (span) => span.data.name.startsWith("task") && span.data.parent_span_id === caseSpan.data.span_id
-    );
-    if (taskSpans.length > 0) {
-      const taskSpan = taskSpans[0];
-      const chatSpans = spans.filter(
-        (span) => span.data.name.startsWith("chat") && span.data.parent_span_id === taskSpan.data.span_id
-      );
-      const chatData = chatSpans.map((chatSpan) => ({
-        operation: chatSpan.data.attributes.custom?.operation || "",
-        capability: chatSpan.data.attributes.custom?.capability || "",
-        step: chatSpan.data.attributes.custom?.step || "",
-        request: {
-          max_token: chatSpan.data.attributes.custom?.["request.max_token"] || "",
-          model: chatSpan.data.attributes.custom?.["request.model"] || "",
-          temperature: chatSpan.data.attributes.custom?.["request.temperature"] || 0
-        },
-        response: {
-          finish_reasons: chatSpan.data.attributes.custom?.["response.finish_reasons"] || ""
-        },
-        usage: {
-          input_tokens: chatSpan.data.attributes.gen_ai?.usage?.input_tokens || 0,
-          output_tokens: chatSpan.data.attributes.gen_ai?.usage?.output_tokens || 0
-        }
-      }));
-      const taskData = {
-        name: taskSpan.data.name,
-        output: taskSpan.data.attributes.custom?.output || "",
-        trial: taskSpan.data.attributes.custom?.trial || 0,
-        type: taskSpan.data.attributes.custom?.type || "",
-        error: taskSpan.data.attributes.custom?.error,
-        chat: chatData[0] || {
-          operation: "",
-          capability: "",
-          step: "",
-          request: { max_token: "", model: "", temperature: 0 },
-          response: { finish_reasons: "" },
-          usage: { input_tokens: 0, output_tokens: 0 }
-        }
-      };
-      caseData.task = taskData;
-    }
-    const scoreSpans = spans.filter(
-      (span) => span.data.attributes.gen_ai.operation.name === "eval.score" && span.data.parent_span_id === caseSpan.data.span_id
-    );
-    caseData.scores = {};
-    scoreSpans.forEach((score) => {
-      const name = score.data.attributes.custom[Attr.Eval.Score.Name];
-      caseData.scores[name] = {
-        name,
-        value: score.data.attributes.custom[Attr.Eval.Score.Value],
-        metadata: {
-          error: score.data.attributes.error
-        }
-      };
-    });
-    rootSpan.cases.push(caseData);
-  }
-  rootSpan.cases.sort((a2, b) => a2.index - b.index);
-  return rootSpan;
-};
-
 // src/util/deep-equal.ts
 function deepEqual(data, other) {
   if (data === other) {
@@ -1383,6 +915,22 @@ function printFinalReport({
   }
 }
 
+// src/config/resolver.ts
+var buildConsoleUrl = (urlString) => {
+  const url = new URL(urlString);
+  return `${url.protocol}//app.${url.host.split("api.").at(-1)}`;
+};
+function resolveAxiomConnection(config) {
+  const consoleEndpointUrl = getConsoleUrl() ?? buildConsoleUrl(config.eval.url);
+  return {
+    url: config.eval.url,
+    consoleEndpointUrl,
+    token: config.eval.token,
+    dataset: config.eval.dataset,
+    orgId: config.eval.orgId
+  };
+}
+
 // src/evals/reporter.ts
 var AxiomReporter = class {
   constructor() {
@@ -1390,7 +938,6 @@ var AxiomReporter = class {
     __publicField(this, "start", 0);
     __publicField(this, "_endOfRunConfigEnd");
     __publicField(this, "_suiteData", []);
-    __publicField(this, "_baselines", /* @__PURE__ */ new Map());
     __publicField(this, "_printedFlagOverrides", false);
     __publicField(this, "_config");
   }
@@ -1415,17 +962,6 @@ var AxiomReporter = class {
       }
       this._printedFlagOverrides = true;
     }
-    const baseline = meta.evaluation.baseline;
-    if (baseline) {
-      const config = getAxiomConfig();
-      if (!config) {
-        throw new AxiomCLIError("Axiom config not available in reporter");
-      }
-      const baselineData = await findEvaluationCases(baseline.id, config);
-      this._baselines.set(meta.evaluation.name, baselineData || null);
-    } else {
-      this._baselines.set(meta.evaluation.name, null);
-    }
     if (meta.evaluation.configEnd && !this._endOfRunConfigEnd) {
       this._endOfRunConfigEnd = meta.evaluation.configEnd;
     }
@@ -1455,16 +991,7 @@ var AxiomReporter = class {
     }
     const cwd = process.cwd();
     const relativePath = testSuite.module.moduleId.replace(cwd, "").replace(/^\//, "");
-    let suiteBaseline = this._baselines.get(meta.evaluation.name);
-    if (suiteBaseline === void 0 && meta.evaluation.baseline) {
-      const config = getAxiomConfig();
-      if (!config) {
-        throw new AxiomCLIError("Axiom config not available in reporter");
-      }
-      const baselineData = await findEvaluationCases(meta.evaluation.baseline.id, config);
-      suiteBaseline = baselineData || null;
-      this._baselines.set(meta.evaluation.name, suiteBaseline);
-    }
+    let suiteBaseline = meta.evaluation.baseline;
     this._suiteData.push({
       name: meta.evaluation.name,
       file: relativePath,
@@ -1533,6 +1060,23 @@ var import_resources = require("@opentelemetry/resources");
 var import_exporter_trace_otlp_http = require("@opentelemetry/exporter-trace-otlp-http");
 var import_api10 = require("@opentelemetry/api");
 
+// src/cli/errors.ts
+var AxiomCLIError = class extends Error {
+  constructor(message) {
+    super(message);
+    this.name = "AxiomCLIError";
+  }
+};
+function errorToString(error) {
+  if (typeof error === "string") {
+    return error;
+  }
+  if (error instanceof Error) {
+    return error.message;
+  }
+  return JSON.stringify(error);
+}
+
 // src/config/loader.ts
 var import_c12 = require("c12");
 var import_defu = require("defu");
@@ -1998,11 +1542,11 @@ function setupEvalProvider(connection) {
   axiomProvider = new import_sdk_trace_node.NodeTracerProvider({
     resource: (0, import_resources.resourceFromAttributes)({
       ["service.name"]: "axiom",
-      ["service.version"]: "0.27.0"
+      ["service.version"]: "0.29.0"
     }),
     spanProcessors: [processor]
   });
-  axiomTracer = axiomProvider.getTracer("axiom", "0.27.0");
+  axiomTracer = axiomProvider.getTracer("axiom", "0.29.0");
 }
 async function initInstrumentation(config) {
   if (initialized) {
@@ -2014,7 +1558,7 @@ async function initInstrumentation(config) {
   }
   initializationPromise = (async () => {
     if (!config.enabled) {
-      axiomTracer = import_api10.trace.getTracer("axiom", "0.27.0");
+      axiomTracer = import_api10.trace.getTracer("axiom", "0.29.0");
       initialized = true;
       return;
     }
@@ -2228,6 +1772,10 @@ function isGlob(str) {
 
 // src/cli/commands/eval.command.ts
 var createRunId = (0, import_nanoid.customAlphabet)("0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ", 10);
+var consoleUrl;
+function getConsoleUrl() {
+  return consoleUrl;
+}
 function getDefaultToken(value) {
   if (typeof value === "string") {
     return value;
@@ -2256,7 +1804,7 @@ var loadEvalCommand = (program2, flagOverrides = {}) => {
         ".",
         "any *.eval.ts file in current directory"
       )
-    ).option("-w, --watch true", "keep server running and watch for changes", false).option("-t, --token <TOKEN>", "axiom token", getDefaultToken).option("-d, --dataset <DATASET>", "axiom dataset name", process.env.AXIOM_DATASET).option("-u, --url <AXIOM URL>", "axiom url", getDefaultUrl).option("-o, --org-id <ORG ID>", "axiom organization id", getDefaultOrgId).option("-b, --baseline <BASELINE ID>", "id of baseline evaluation to compare against").option("--debug", "run locally without sending to Axiom or loading baselines", false).option("--list", "list evaluations and test cases without running them", false).action(async (target, options) => {
+    ).option("-w, --watch true", "keep server running and watch for changes", false).option("-t, --token <TOKEN>", "axiom token", getDefaultToken).option("-d, --dataset <DATASET>", "axiom dataset name", process.env.AXIOM_DATASET).option("-u, --url <AXIOM URL>", "axiom url", getDefaultUrl).option("-o, --org-id <ORG ID>", "axiom organization id", getDefaultOrgId).option("-b, --baseline <BASELINE ID>", "id of baseline evaluation to compare against").option("--debug", "run locally without sending to Axiom or loading baselines", false).option("--list", "list evaluations and test cases without running them", false).addOption(new import_commander.Option("-c, --console-url <URL>", "console url override").hideHelp()).action(async (target, options) => {
       try {
         if (options.debug) {
           process.env.AXIOM_DEBUG = "true";
@@ -2301,6 +1849,7 @@ var loadEvalCommand = (program2, flagOverrides = {}) => {
           console.log("");
         }
         const runId = createRunId();
+        consoleUrl = options.consoleUrl;
         await runEvalWithContext(flagOverrides, async () => {
           return runVitest(".", {
             watch: options.watch,
@@ -2786,7 +2335,7 @@ var import_commander2 = require("commander");
 var loadVersionCommand = (program2) => {
   return program2.addCommand(
     new import_commander2.Command("version").description("cli version").action(() => {
-      console.log("0.27.0");
+      console.log("0.29.0");
     })
   );
 };
@@ -2796,7 +2345,7 @@ var { loadEnvConfig } = import_env.default;
 loadEnvConfig(process.cwd());
 var { cleanedArgv, overrides } = extractOverrides(process.argv.slice(2));
 var program = new import_commander3.Command();
-program.name("axiom").description("Axiom's CLI to manage your objects and run evals").version("0.27.0");
+program.name("axiom").description("Axiom's CLI to manage your objects and run evals").version("0.29.0");
 program.hook("preAction", async (_, actionCommand) => {
   const commandName = actionCommand.name();
   const parentCommand = actionCommand.parent;