FinceptTerminal is a modern finance application offering advanced market analytics, investment research, and economic data tools, designed for interactive exploration and data-driven decision-making in a user-friendly environment.
C++ ai-agentsalgorithmic-tradingbloomberg-terminalcppfinancefinancial-marketsfintechgood-first-issueinvestmentinvestment-researchmachine-learningopensourcepythonqtquantitative-financestock-markettrading
// Per-provider model-list discovery. Async GET on the GUI thread; emits models_fetched.
#include "auth/AuthManager.h"
#include "core/config/AppConfig.h"
#include "core/logging/Logger.h"
#include "services/llm/LlmContentExtractors.h"
#include "services/llm/LlmService.h"
#include "services/llm/ProviderCatalog.h"
#include "storage/repositories/SettingsRepository.h"
#include <QByteArray>
#include <QJsonArray>
#include <QJsonDocument>
#include <QJsonObject>
#include <QJsonValue>
#include <QNetworkAccessManager>
#include <QNetworkReply>
#include <QNetworkRequest>
#include <QRegularExpression>
#include <QString>
#include <QStringList>
#include <QTimer>
#include <QUrl>
namespace fincept::ai_chat {
static constexpr const char* TAG = "LlmModelsApi";
QString LlmService::get_models_url(const QString& provider, const QString& api_key, const QString& base_url) {
Q_UNUSED(api_key) // Auth in headers, not URL.
const QString p = provider.toLower();
// Blocked provider (AtlasCloud, removed) — no models URL, so the Fetch button
// can never reach it even if base_url was hand-pointed at the host.
if (ProviderCatalog::is_blocked(p, base_url))
return {};
// Ollama uses /api/tags, not /v1/models — must short-circuit before the custom-base_url branch
// or a user with base_url=http://localhost:11434 hits the wrong path and parses a 404 page.
if (p == "ollama") {
QString base = base_url.isEmpty() ? "http://localhost:11434" : base_url;
if (base.endsWith('/'))
base.chop(1);
return base + "/api/tags";
}
// Custom base_url (except fincept) — assume OpenAI-compatible /v1/models.
if (!base_url.isEmpty() && p != "fincept") {
QString base = base_url;
while (base.endsWith('/'))
base.chop(1);
const QString suffix = (p == "anthropic") ? QStringLiteral("/models?limit=1000") : QStringLiteral("/models");
// Already a full endpoint — use verbatim.
if (base.contains(QStringLiteral("/models")))
return base;
// Base already includes a version segment (e.g. ".../v1", ".../v1beta") —
// append only the suffix. Otherwise inject the default "/v1".
static const QRegularExpression re(QStringLiteral("/v\\d+[a-zA-Z]*$"));
if (re.match(base).hasMatch())
return base + suffix;
return base + QStringLiteral("/v1") + suffix;
}
if (p == "openai")
return "https://api.openai.com/v1/models";
if (p == "anthropic")
return "https://api.anthropic.com/v1/models?limit=1000";
if (p == "gemini" || p == "google") {
return "https://generativelanguage.googleapis.com/v1beta/models?pageSize=100";
}
if (p == "groq")
return "https://api.groq.com/openai/v1/models";
if (p == "deepseek")
return "https://api.deepseek.com/models";
if (p == "openrouter")
return "https://openrouter.ai/api/v1/models";
if (p == "xai")
return "https://api.x.ai/v1/models";
if (p == "kimi")
return "https://api.moonshot.ai/v1/models";
if (p == "aihubmix")
return "https://aihubmix.com/v1/models"; // fallback if prefilled base_url was cleared
// fincept publishes no models endpoint — /research/llm/models is a 404, and
// /research/llm/async takes no `model` field at all (the backend picks).
// fetch_models() short-circuits to the known list before reaching here.
// minimax has no public /v1/models — caller falls back to known models.
return {};
}
QMap<QString, QString> LlmService::get_models_headers(const QString& provider, const QString& api_key) {
QMap<QString, QString> h;
const QString p = provider.toLower();
if (p == "anthropic") {
if (!api_key.isEmpty())
h["x-api-key"] = api_key;
h["anthropic-version"] = "2023-06-01";
} else if (p == "gemini" || p == "google") {
if (!api_key.isEmpty())
h["x-goog-api-key"] = api_key;
} else if (p == "ollama") {
// No auth.
} else if (p == "fincept") {
// Same fallback as ensure_config — resolve via AuthManager
// (session → SecureStorage), never the plaintext settings row (CR-08).
QString resolved_key = api_key;
if (resolved_key.isEmpty())
resolved_key = fincept::auth::AuthManager::instance().fincept_api_key();
if (!resolved_key.isEmpty())
h["X-API-Key"] = resolved_key;
} else {
// OpenAI-compatible.
if (!api_key.isEmpty())
h["Authorization"] = "Bearer " + api_key;
}
return h;
}
QStringList LlmService::parse_models_response(const QString& provider, const QByteArray& body) {
QStringList models;
auto doc = QJsonDocument::fromJson(body);
if (doc.isNull() || !doc.isObject())
return models;
QJsonObject root = doc.object();
const QString p = provider.toLower();
if (p == "gemini" || p == "google") {
QJsonArray arr = root["models"].toArray();
for (const auto& v : arr) {
QJsonObject m = v.toObject();
// Filter to models that support generateContent.
QJsonArray methods = m["supportedGenerationMethods"].toArray();
bool can_generate = false;
for (const auto& method : methods) {
if (method.toString() == "generateContent") {
can_generate = true;
break;
}
}
if (!can_generate)
continue;
QString name = m["name"].toString();
if (name.startsWith("models/"))
name = name.mid(7);
if (!name.isEmpty())
models.append(name);
}
} else if (p == "ollama") {
QJsonArray arr = root["models"].toArray();
for (const auto& v : arr) {
QString name = v.toObject()["name"].toString();
if (!name.isEmpty())
models.append(name);
}
} else if (p == "fincept") {
// data may be either {"models":[...]} or a direct array.
QJsonValue data_val = root["data"];
QJsonArray arr;
if (data_val.isObject())
arr = data_val.toObject()["models"].toArray();
else if (data_val.isArray())
arr = data_val.toArray();
for (const auto& v : arr) {
QString id = v.isString() ? v.toString() : v.toObject()["id"].toString();
if (!id.isEmpty())
models.append(id);
}
if (models.isEmpty())
models = {"MiniMax-M2.7", "MiniMax-M2.7-highspeed", "MiniMax-M2.5"};
} else {
// OpenAI-compatible: {"data": [{"id": ...}]}.
QJsonArray arr = root["data"].toArray();
for (const auto& v : arr) {
QString id = v.toObject()["id"].toString();
if (!id.isEmpty())
models.append(id);
}
}
models.sort(Qt::CaseInsensitive);
return models;
}
void LlmService::fetch_models(const QString& provider, const QString& api_key, const QString& base_url) {
// No public /models endpoint for fincept — return known list immediately.
if (provider.toLower() == "fincept") {
emit models_fetched(provider,
{"MiniMax-M2.7", "MiniMax-M2.7-highspeed", "MiniMax-M2.5", "MiniMax-M2.5-highspeed"}, {});
return;
}
const QString url = get_models_url(provider, api_key, base_url);
if (url.isEmpty()) {
emit models_fetched(provider, {}, "Unknown provider: " + provider);
return;
}
// NAM must live on the main thread — its DNS/TLS path needs a pumping event loop.
if (!models_nam_)
models_nam_ = new QNetworkAccessManager(this);
QNetworkRequest req{QUrl(url)};
req.setAttribute(QNetworkRequest::RedirectPolicyAttribute, QNetworkRequest::NoLessSafeRedirectPolicy);
auto headers = get_models_headers(provider, api_key);
for (auto it = headers.constBegin(); it != headers.constEnd(); ++it)
req.setRawHeader(it.key().toUtf8(), it.value().toUtf8());
LOG_INFO(TAG, QString("fetch_models start: provider=%1 url=%2").arg(provider, url));
QNetworkReply* reply = models_nam_->get(req);
reply->setProperty("_provider", provider);
auto* timer = new QTimer(reply);
timer->setSingleShot(true);
connect(timer, &QTimer::timeout, reply, [reply]() {
if (reply->isRunning())
reply->abort();
});
timer->start(15000);
connect(reply, &QNetworkReply::finished, this, [this, reply]() {
reply->deleteLater();
const QString p = reply->property("_provider").toString();
QString error;
QStringList models;
const int status = reply->attribute(QNetworkRequest::HttpStatusCodeAttribute).toInt();
if (reply->error() != QNetworkReply::NoError || status < 200 || status >= 300) {
if (reply->error() == QNetworkReply::OperationCanceledError)
error = "Request timed out";
else {
error = reply->errorString();
if (error.isEmpty())
error = QString("HTTP %1").arg(status);
}
// Logged before the provider's message is folded in: that text is for the
// user (it can echo a masked key fragment), not for the log file.
LOG_WARN(TAG, QString("fetch_models %1 failed: %2 (status=%3 qt_error=%4)")
.arg(p, error)
.arg(status)
.arg(static_cast<int>(reply->error())));
// Qt's "server replied: Unauthorized" says nothing about WHY — the provider's
// JSON body usually does ("API key not valid", "model access denied", …).
if (reply->error() != QNetworkReply::OperationCanceledError) {
const QString server_msg = parse_server_error_message(reply->readAll());
if (!server_msg.isEmpty())
error = status > 0 ? QString("HTTP %1: %2").arg(status).arg(server_msg) : server_msg;
}
} else {
models = parse_models_response(p, reply->readAll());
if (models.isEmpty())
error = "No models found";
LOG_INFO(TAG, QString("fetch_models %1 → %2 models").arg(p).arg(models.size()));
}
emit models_fetched(p, models, error);
});
}
} // namespace fincept::ai_chat